agent based modeling of power distribution systems

Graduate Theses, Dissertations, and Problem Reports

2009

Agent based modeling of power distribution systems Agent based modeling of power distribution systems

Sridhar Chouhan West Virginia University

Follow this and additional works at: https://researchrepository.wvu.edu/etd

Recommended Citation Recommended Citation Chouhan, Sridhar, "Agent based modeling of power distribution systems" (2009). Graduate Theses, Dissertations, and Problem Reports. 1996. https://researchrepository.wvu.edu/etd/1996

This Thesis is protected by copyright and/or related rights. It has been brought to you by the The Research Repository @ WVU with permission from the rights-holder(s). You are free to use this Thesis in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you must obtain permission from the rights-holder(s) directly, unless additional rights are indicated by a Creative Commons license in the record and/ or on the work itself. This Thesis has been accepted for inclusion in WVU Graduate Theses, Dissertations, and Problem Reports collection by an authorized administrator of The Research Repository @ WVU. For more information, please contact [email protected].

https://researchrepository.wvu.edu/

https://researchrepository.wvu.edu/

https://researchrepository.wvu.edu/etd

https://researchrepository.wvu.edu/etd?utm_source=researchrepository.wvu.edu%2Fetd%2F1996&utm_medium=PDF&utm_campaign=PDFCoverPages

https://researchrepository.wvu.edu/etd/1996?utm_source=researchrepository.wvu.edu%2Fetd%2F1996&utm_medium=PDF&utm_campaign=PDFCoverPages

mailto:[email protected]

Agent Based Modeling of

Power Distribution Systems

by

Sridhar Chouhan

Thesis submitted to the College of Engineering and Mineral Resources

at West Virginia University in partial fulfillment of the requirements

for the degree of

Master of Science

in

Electrical Engineering

Prof. Ali Feliachi, Ph.D., Chair Prof. Daryl S Reynolds, Ph.D.

Prof. Muhammad A. Choudhry, Ph.D.

Lane Department of Computer Science and Electrical Engineering

Morgantown, West Virginia 2009

Keywords: Multi-Agent System, Fault Detection, Reconfiguration, DEW©, JADE, Power World Simulator.

Copyright 2009 Sridhar Chouhan

ABSTRACT

Agent Based Modeling of Power Distribution Systems by

Sridhar Chouhan

Master of Science in Electrical Engineering

West Virginia University

Professor Ali Feliachi, Ph.D., Chair

The electric power system is a very vast network and becoming more complex each day. The traditional vertically monopolistic structure has been deregulated and replaced by gencos, transcos and, discos; increasing the power system intricacy. During the past few decades there has been remarkable development in software and hardware technologies for the analysis and design activities in power system planning, operation, and control. However, much still depends on the judgment of human experts. A single fault in power system can lead to multiple faults and can collapse the whole system. Power System needs a more decentralized control mechanism for solving these problems. One novel solution would be Multi-agent Systems. A Multi-agent system is a collection of agents, which perceives the system changes and acts on the system in order to achieve its goals. Recent technology developments in the area of Multi-agent systems making it a viable solution for today’s complicated power network. A Multi-agent system model is developed for fault detection and reconfiguration in this thesis work. These models are developed based on graph theory tree models and mathematical models. A set of objective functions are specified in the mathematical model for the restoration of the network. The agent platform for the fault detection is developed by Java Agent Development Framework. The restoration algorithm is programmed in MATLAB and applied to the distribution system modeled in the commercial software, Distributed Engineering Workstation and Power World Simulator. The test system in this thesis is, a distribution system developed by Southern California Edison called Circuit of the Future. The Multi-agent system can detect the fault precisely and reconfigures the circuit using the reconfiguration algorithm. The reconfiguration will happen in a way that it always try to supply all the critical loads in the network. When there are multiple solutions available for reconfiguration, the one with good voltage profile and less power loss is selected as the solution. The algorithm makes use of shunt compensation and priority based load shedding in order to control the voltage across the network. Agents make use of learning to speed up the reconfiguration process.

iii

DEDICATION

I would like to dedicate this thesis to my parents,

who carved me to whatever little I am today,

&

My adorable cousin, Koneri Srinivas,

who helped me in each mode of my life.

***

Grand pa, you are deeply missed!

iv

The only living gods on earth are,

‘Parents’

v

ACKNOWLEDGEMENTS

After two years of my degree, I have learned one thing – I could never have completed my

research work and the thesis dissertation, without the support and encouragement of

many people around me. First of all, I would like to thank my esteemed mentor and

advisor, Dr. Ali Feliachi. I owe him so much. He has given me a chance to continue my

work in the field of power systems, which is actually my passion. He taught me much of

what I know today and guided me throughout my research work. I would also like to thank

rest of my thesis committee, Dr. Muhammad A Choudhry and Dr. Daryl S Reynolds, for

their support and advice. I would like to thank Dr. Hong-Jian Lai for helping me out with

the mathematical modeling used in this thesis work.

I extend the appreciation to Professor Hui Wan, all other faculty members in APERC and

the staff in LDCSEE for their encouragement and support. I would also like to express my

thanks to my friends in APERC for the healthy discussions regarding my research work. A

special thanks to Ms. Koushaly Nareshkumar for helping me in understanding Multi-agent

concepts related to power distribution systems. For assistance in JADE programming, my

thanks are extended to Ms. Summiya Moheuddin. I’ve been fortunate enough to have a

great group of friends at WVU. I would like to thank all of them for making my master’s life

joyful and memorable. Specially, I would like to thank my best friends Usha, Navya, and

all my roommates, who were always there for me when I needed moral support.

I would like to thank my family for their unending support and love from childhood to now.

Without their support I couldn’t have made it through this process or any of the tough

times in my life. My sincere gratitude goes to my cousin for helping and motivating me

throughout my life. My humble pranam goes to the almighty god for being with me as a

guiding light throughout my life.

This work was supported in part by grants from the US DEPSCoR/ONR grant No.

N00014-03-1-0660 and the US DoE grant No. DE-FC26-06NT42793.

vi

CONTENTS

ABSTRACT ..........................................................................................................................ii

Dedication ...........................................................................................................................iii

Acknowledgements ............................................................................................................. v

Contents..............................................................................................................................vi

List of Figures...................................................................................................................... x

List of Tables.....................................................................................................................xiii

Nomenclature....................................................................................................................xiv

Chapter 1 ............................................................................................................................ 1

INTRODUCTION................................................................................................................. 1

1.1 Background ......................................................................................................... 1

1.2 Multi-Agent Systems............................................................................................ 3

1.3 Agent Communication ......................................................................................... 7

1.3.1 FIPA-ACL ........................................................................................................... 8

1.3.2 Knowledge Query and Manipulation Language (KQML) .................................. 10

1.4 Problem Statement............................................................................................ 11

1.5 Approach ........................................................................................................... 12

1.6 Outline ............................................................................................................... 13

Chapter 2 .......................................................................................................................... 14

vii

Literature Review .............................................................................................................. 14

2.1 Multi-agent systems........................................................................................... 14

2.2 MAS in Automatic Fault Detection and Reconfiguration.................................... 16

2.3 MAS in Distribution System with DER ............................................................... 20

2.4 MAS in protection coordination.......................................................................... 21

2.5 MAS in Voltage Control ..................................................................................... 24

2.6 MAS in Power Markets ...................................................................................... 25

2.7 MAS in other engineering disciplines ................................................................ 26

2.8 Limitations and Technical challenges of MAS ................................................... 27

Chapter 3 .......................................................................................................................... 30

Simulation Software .......................................................................................................... 30

3.1 MATLAB ®......................................................................................................... 31

3.2 MATPOWER ..................................................................................................... 32

3.3 Java Agent Development Framework ............................................................... 32

3.3.1 JADE Architecture ............................................................................................ 33

3.3.2 JADE Programming.......................................................................................... 37

3.4 Distributed Engineering Workstation© (DEW)................................................... 37

3.5 Power World Simulator...................................................................................... 39

Chapter 4 .......................................................................................................................... 42

Mathematical Model and Algorithms for MAS ................................................................... 42

4.1 Mathematical Model .......................................................................................... 42

viii

4.1.1 Graph Theory ................................................................................................... 42

4.1.2 Graph Theory representation of Power System ............................................... 43

4.2 Fault Detection Algorithm .................................................................................. 45

4.3 Fault Reconfiguration & Restoration Algorithm ................................................. 45

4.3.1 Reconfiguration algorithm for single fault ......................................................... 46

4.3.2 Reconfiguration algorithm for two faults ........................................................... 46

4.3.3 Reconfiguration algorithm for three faults ........................................................ 46

4.3.4 Objective functions and Constraints................................................................. 48

4.4 Illustration .......................................................................................................... 49

Chapter 5 .......................................................................................................................... 51

Simulation and Results...................................................................................................... 51

5.1 Proposed MAS Architecture .............................................................................. 51

5.2 Circuit of the Future (CoF)................................................................................. 54

5.3 Fault Detection and Reconfiguration Algorithm ................................................. 57

5.3.1 Test Case 1: For a Single Fault........................................................................ 59

5.3.2 Test Case 2: For Multiple Faults ...................................................................... 63

5.4 MATLAB Reconfiguration .................................................................................. 65

5.4.1 Test Case 3: For a Line Fault ........................................................................... 68

5.4.2 Test Case 4: For two Line Faults ..................................................................... 73

5.4.3 Test Case 5: For three Line Faults ................................................................... 76

5.4.4 Test Case 6: Special Scenario ......................................................................... 78

ix

5.4.5 Test Case 7: Special Scenario ......................................................................... 80

5.5 Algorithm Execution Duration ............................................................................ 82

Chapter 6 .......................................................................................................................... 84

Conclusion and Future Work............................................................................................. 84

6.1 Conclusion......................................................................................................... 84

6.2 Future Work....................................................................................................... 86

References........................................................................................................................ 88

x

LIST OF FIGURES

Figure 1-1: A Multi-agent System for Power System. ......................................................... 4

Figure 1-2: FIPA ACL message [14]. .................................................................................. 9

Figure 1-3: KQML communication protocol [6]................................................................. 11

Figure 3-1: Architecture of the Simulation Model. ............................................................. 31

Figure 3-2: Relationship between main architectural elements [67].................................. 34

Figure 3-3: JADE Graphical User Interface (GUI). ............................................................ 36

Figure 3-4: JADE run-time environment............................................................................ 36

Figure 3-5: Organizational structure of DEW [69]. ............................................................ 38

Figure 3-6: Distributed Engineering Workstation’s Working Environment......................... 39

Figure 3-7: Snapshot of Circuit of the Future one line diagram......................................... 40

Figure 3-8: Snapshot of Model Explorer of Power World Simulator [70]........................... 41

Figure 4-1: Graph model of power distribution system...................................................... 44

Figure 4-2: Flow Chart for Fault Reconfiguration .............................................................. 47

Figure 4-3: Representation of Simplified CoF. .................................................................. 50

Figure 5-1: Proposed Multi-agent system architecture...................................................... 52

Figure 5-2: Global Agent (GAG) functionalities. ................................................................ 53

Figure 5-3: The Circuit of the Future modeled in DEW [2]. ............................................... 55

Figure 5-4: The Simplified CoF modeled in Power World Simulator. ................................ 56

Figure 5-5: Flow chart for Fault Detection and Reconfiguration........................................ 58

Figure 5-6: Simplified CoF with fault in line 2-3................................................................. 59

Figure 5-7: Power World Simulator’s Load flow report for line fault at line 2-3. ................ 60

Figure 5-8: JADE message passing between agents for line fault at line 2-3. .................. 60

xi

Figure 5-9: Java output console showing results of Test Case 1...................................... 61

Figure 5-10: MATLAB plot depicting reconfiguration network for fault at line 2-3. ............ 62

Figure 5-11: JADE message passing between agents for line faults at line 1-2 and 1-5. . 63

Figure 5-12: Java output console showing results of Test Case 2.................................... 63

Figure 5-13: MATLAB plot depicting reconfiguration network for fault at line 1-2 and 1-5.64

Figure 5-14: MATLAB GUI for Fault Reconfiguration........................................................ 65

Figure 5-15: MATLAB plot depicting basic Simplified CoF................................................ 67

Figure 5-16: MATLAB plot depicting voltage profile with S1 closed for line fault at 1-2..... 68



Figure 5-19: MATLAB plot depicting voltage profile with S1&S5 closed for line fault at 1-2.

.......................................................................................................................................... 70


.......................................................................................................................................... 71

Figure 5-21: MATLAB plot depicting reconfiguration network for fault at line 1-2. ............ 72

Figure 5-22: Plot depicting voltage profile with S1, S5 & S7 closed for line faults at 1-2 &

14-15, without voltage control. .......................................................................................... 73

Figure 5-23: Plot depicting voltage profile with S1, S5 & S7 closed for line faults at 1-2 &

14-15 with shunt compensation......................................................................................... 74

Figure 5-24: MATLAB plot depicting reconfiguration network for fault at lines 1-2 and 14-

15. ..................................................................................................................................... 75

Figure 5-25: Plot depicting voltage profile with S1, S5 & S7 closed for line faults at 1-2, 14-

15 ...................................................................................................................................... 76

Figure 5-26: MATLAB plot depicting reconfiguration network for fault at lines 1-2, 14-15

and 11-12. ......................................................................................................................... 77

Figure 5-27: MATLAB output command window showing solution of Test Case 5........... 78

xii

Figure 5-28: MATLAB plot depicting reconfiguration network for fault at lines 9-10 and 10-

11. ..................................................................................................................................... 79

Figure 5-29: MATLAB plot depicting reconfiguration network for fault at line 1-2……… 81 Figure 5-30: Plot of Execution Time for Fault Detection Algorithm [2]. ............................. 82

Figure 5-31: Plot of Execution Time for Fault Reconfiguration Algorithm. ........................ 82

Figure 5-32: Plot of Execution Time for Fault Reconfiguration Algorithm accessing data

base. ................................................................................................................................. 83

xiii

LIST OF TABLES

Table 1-1: FIPA ACL Message Parameters [14]. .............................................................. 10

Table 2-1: SWOT matrix for current Agent Applications to power systems [15]. .............. 15

Table 4-1: Description of the notations used in Mathematical model................................ 43

xiv

NOMENCLATURE

ACL Agent Communication Language

CoF Circuit of the Future

DER Distributed Energy Resources

DEW Distributed Engineering Workstation

DG Distributed Generators

FIPA Foundation for Intelligent Physical Agents

GUI Graphical User Interface

JADE Java Agent Development Framework

KQML Knowledge Query and Manipulation Language

LAG Load Agent

GAG Global Agent

MAS Multi-agent System

OOP Object Oriented Programming

SAG Switch Agent

SCE Southern California Edison

WVU West Virginia University

1

Chapter 1

INTRODUCTION

1.1 Background

The Electrical Power System is the most capital intensive and most complex system ever

developed by man. The main era of electrical power system was started by the invention

of electric bulb by Thomas Alva Edison in 1879. He also started generating the electricity

by building small generating stations. Edison’s method of generating and transmitting

electricity was called Direct Current (DC) or Low Voltage. The main problem of DC power

was that it cannot be transmitted to long distances and cannot be transformed from one

voltage level to another. During this time all the loads were supplied at the same voltage

level [1].

In 1888, Nikola Tesla invented the alternating current generation. In an Alternating Current

(AC) system, transformers were used to step up the voltage levels and enabled the

electricity to transfer over long distances. Starting from then many forms of generating the

electricity came in to existence like, Thermal Power Generation, Hydro Power Generation,

Diesel Power Generation and Gas Power Generations. Development in Nuclear Physics

started the nuclear power generation in the recent years and made it a cheaper way of

generating power. Because of the technology development even renewable energy

sources are making their mark on the power generation.

Chapter 1: Introduction

2

Today Distributed Energy Resources (DERs), some times called Distributed Generators

(DGs) are provided at the distribution level for serving some important local loads such as,

hospital loads. The main reason for providing DER is that it serves as a local generation

when there is a shortage of power supply through the grid. Power system is suffering from

congestion issues because of deregulation. These issues can be solved by installing

DERs near the load centers. However, Interconnection of these DERs is making the

system more complicated. After inserting the DERs in to the system, we need to make

sure that the system stability and voltage profile are within the safe limits. Even we need

to come up with new protection coordination schemes to incorporate these DERs in to the

system [2].

The Electric Power System is growing rapidly and becoming more complex each day

because of its continuously increasing power demand. Besides the increasing demand

deregulation in the power industry is another fact, which is making today’s electric power

system to be more intrinsic. In the past, the electric power industry has been vertically

integrated i.e., a central authority monitored and controlled all the activities in generation,

transmission, and distribution. For the last decade or so, the electric power industry has

been undergoing a process of restructuring, especially separation of transmission from

generation activities. Deregulation is intended to encourage competition among utilities

and power marketers to reduce energy prices. There is a need of a power system

operator, to coordinate the dispatch of generating units to meet the expected demand of

the system across the transmission grid. However this deregulation brought so much of

complexity in the electric power system which is needed to be handled carefully [3].

Electric power system is such a vast and sensitive network that even a single fault can

cause the whole system to be down. A single fault normally causes another fault in the

system and this process keeps on going until the whole system collapses. This particular

phenomenon in power system is said to be Cascading Effect. A power system operator,

who generally monitors and controls the whole system, has to be very careful regarding

this cascading phenomenon. A small human mistake can cause a massive black out in

the power network leading to billions of dollar loss. A great example for this type of black

out is the 2003 Northeast Blackout [4]. The major reason for the Northeast Blackout was a

sudden power fluctuation in a particular area of Ohio. The outage of Ohio generating

station triggered the cascading effect, causing several high voltage transmission lines to


3

go over their thermal limit. These sagged transmission lines then touched the overgrown

trees in Ohio area and caused several ground faults. These things cascaded together and

brought down approximately 100 generating stations. This was one of the major blackouts

in the power history which caused nearly 50 million people to live in dark and 6 billion

dollars of financial losses [4].

The one reason behind this massive havoc was lack of controlling on power grid using

EMS (Energy Management Systems) and Restoration techniques available at that time.

The power quality and reliability can be improved by having better fault detection and

reconfiguration techniques. These can be achieved by a novel technique called Multi-

agent systems. Multi-agent systems monitor and control the power systems in an

autonomous and decentralized manner. They use their communication capability to

communicate and negotiate on a particular task. Recent technology developments and

drastic reduction of cost are making this Multi-agent system a most viable solution.

1.2 Multi-Agent Systems

Agent based technologies have been the subject of extensive research and investigation

for several years, but it is recently that these have been seen any significant degree of

exploitation in commercial applications. Multi-agent systems have been used in vivid

applications, ranging from small systems to most intrinsic systems. Some of the fields

where these Multi-agent systems have been fruitfully employed include process control,

system diagnostics, manufacturing, transportation logistics and network management [5].

Before understanding the Multi-agent systems one has to understand the concepts behind

the single agent. Many people have defined the word ‘agent’ in their own ways since the

starting of agent technologies. Out of many definitions available a couple of relevant

definitions are given below

“An agent is an autonomous computational entity such as a software program that can be

viewed as perceiving its environment through sensors and acting upon this environment

through its effectors” – Weiss [6].


4

“An agent is a computer system that is situated in some environment, and that is capable

of autonomous action in this environment in order to meet its design objectives” –

Wooldridge and Jennings (1995) [7].

The above two definitions are much suitable for this thesis project. Multi-agent systems

received greater attention in the recent past in many intelligent, distributed systems to

model the theories of interactivity in human societies. Humans interact in various ways

and at different levels. For instance, humans observe and model one another, they

request and provide information, they negotiate and discuss, they develop shared views of

their environment, they detect and resolve conflicts, and they form organizational

structures such as teams, committees, and economies [8].

In this thesis the term agent means, a software entity which perceives its environmental

changes and acts on it diligently to achieve its organizational objectives. A Multi-agent

system means the collection of such agents which communicate and coordinate to solve a

particular problem according to their objectives. The Multi-agent system approach for

power system is shown as below,

Power System

Senses Act

Agent

Figure 1-1: A Multi-agent System for Power System.


5

The main problem facing by an agent is deciding which of its actions it should perform in

order to best satisfy its organizational objectives. The type of decision selected always

depends upon the type of environment. Russell and Norvig [9] proposed following

classification of environment properties,

• Accessible vs Inaccessible.

An accessible environment is one in which all the states of the environment are

easily accessible. But most of the today’s complex networks, such as every day

physical world and Internet, are inaccessible.

• Deterministic vs Non-deterministic

In deterministic environment the actions will have certain effects on the

environment. In non-deterministic environments it is difficult to anticipate the

effects of particular action on the environment. Non-deterministic environments will

have greater difficulty in designing.

• Episodic vs Non-episodic

In an episodic environment, the performance of an agent is dependent on a

number of episodes, such as mail sorting system.

• Static vs Dynamic

A static environment is one that can be assumed to be unchanged except for the

actions by agent. In contrast to the static environment the real world is most

dynamic one which changes time to time hence, difficult to control.

• Discrete vs Continuous

An environment is said to be discrete, if it has discrete number of actions and

percepts in it. The best example for a discrete environment is chess game and for

a continuous environment is car driving.

Comparing the agents to the present world computer systems, every action a computer

performs is anticipated, planned for, and coded by a programmer. When a computer

encounters a new sort of problem, which is not anticipated, then the system will crash.

This is because the computer systems don’t have intelligence embedded in it. Unlike the

computer systems the agents are intelligent enough to take up the flexible autonomous

actions to satisfy their objectives [7]. Here the flexibility means three things


6

• Reactivity:

Agents are able to perceive their environment, and respond in a timely fashion to

changes that occur in it, in order satisfy their objectives.

• Pro-activeness:

Intelligent agents are able to exhibit goal oriented behavior by taking the initiative

in order to satisfy their objectives.

• Social-ability:

Intelligent agents are able to communicate and cooperate with other agents

available in Multi-agent systems in order to satisfy their objectives.

Out of the above three characteristics the social-ability is more complex one. In the real

world daily millions of computers exchange the information with both humans and other

computers. But the ability to exchange the bit stream is not really the social ability.

Consider the human world, humans achieve their goals by negotiation and cooperation

[6]. The same way intelligent agents cooperate and negotiate with each other to achieve

the common goal of the Multi-agent system. The communication between the agents in

MAS has certain standards defined by FIPA (Foundation for Intelligent Physical Agents).

There is always been a great amount confusion between agents and objects, expert

systems. The main distinctions between traditional view of an object and our view of an

agent is listed below [6],

• Agent embody a strong notion of autonomy than objects, they decide on

themselves whether or not to perform an action on request from other agents.

• Agents are capable of flexible (reactive, pro-active, social) behavior. Where

objects have nothing to do with these behaviors.

• A MAS is inherently multi-threaded, each agent has at least one thread of control.

Expert systems are most important AI (Artificial Intelligence) technology of 1980s [10]. An

Expert system is one that is capable of solving problems or giving advice in some

knowledge rich domain [6] [11]. The main difference between Expert systems and Agent

technologies is that, an Expert system cannot sense its environment directly. Instead, it


7

needs a middle man to input the data. In the same way an Expert system cannot act

directly on its environment. It just gives certain instructions to the operator.

1.3 Agent Communication

It has been seen in the earlier section that agents will communicate to coordinate and

negotiate on particular task to achieve its organizational objectives. The communication

ability of agents is part perception (the receiving of messages) and part action (the

sending of messages). There are generally two message types in Agent Communication

they are, Assertion and Queries. Every agent, whether active or passive, must have the

capability to accept the information. In order to assume a passive role in a dialogue, an

agent must be able to 1) accept a query from an external source and 2) send a reply to

the source by making an assertion. In order to assume an active role in a dialogue, an

agent must have the capability to send queries and make assertions [6].

Communication protocols are very important in any communication systems. The Multi-

agent communication is specified by a data structure with the following five fields,

1. Sender.

2. Receiver(s).

3. Language in protocol.

4. Encoding and Decoding functions.

5. Action to be taken by the receiver(s).

Speech Act Theory helps in defining the message send by sender. So, that the receiver

has no doubts regarding the received message [12]. In order to understand and to be

understood communication messages precisely, all the agents in MAS has to follow the

same ontology. Two agents following the same ontology means, those two agents will

know the meaning of messages during the communication. Hence, ontology results in

perfect operation of communication. In technical terms the name ontology is defined as

follows,


8

“Ontology is a formal definition of a body of knowledge. The most typical type of ontology

used in building agents involves a structural component. Essentially a taxonomy of class

and subclass relations coupled with definitions of the relationships between these things”

– Jim Hendler [6]

The main languages which are followed for the agent communication are,

• Agent Communication Language (ACL) by the Foundation for Intelligent Physical

Agents (FIPA)

• Knowledge Query and Manipulation Language (KQML)

1.3.1 FIPA-ACL

Foundation for Intelligent Physical Agents (FIPA) is an IEEE Computer Society Standards

organization that promotes agent-based technology and the interoperability of its

standards with other technologies. FIPA was originally formed as a Swiss based

organization in 1996 to produce software standards specifications for heterogeneous and

interacting agents and agent based systems [13].

FIPA has developed a certain number of standards for heterogeneous agent

communications. FIPA Agent Communication specifications deal with Agent

Communication Language (ACL) messages, message exchange interaction protocols,

speech act theory based communicative acts, and content language representations.

Agent Communication Language (ACL) developed by FIPA is based on speech act

theory. Here the messages are actions, or communicative acts, as they are intended to

perform some action by virtue of being sent. The FIPA ACL specifications consist of a set

of message types and the description of their pragmatics, which is the effect on the mental

attitudes of sender and receiver agents. A sample FIPA ACL message is as shown below,


9

As noted above, the message contains many parameters. These parameters can occur in

any order in the message. The only parameter which is mandatory for the agent

communication is :receiver parameter. However, for a meaningful message there should

be other parameters required depending upon the situation. The full set of pre-defined

message parameters is shown in the following table.

Message Parameter

Meaning:

:sender Denotes the identity of the sender of the message. i.e., the name of the

agent.

:receiver Denotes the identity of the intended recipient of the message.

:content Denotes content of the message; equivalently denotes the object of the

action.

:reply-with An expression which will be used by the agent responding to this

message to identify the original message. Can be used to follow a

Conversation thread in a situation where multiple dialogues occur

simultaneously.

:in-reply-to Denotes an expression that references an earlier action to which this

message is a reply.

:envelope Denotes an expression that provides useful information about the

Figure 1-2: FIPA ACL message [14].


10

message as seen by the message transport service. The content of this

parameter is not defined in the specification, but may include time sent,

time received, route, etc.

:language Denotes the encoding scheme of the content of the action.

:ontology Denotes the ontology which is used to give a meaning to the symbols

in the content expression.

:reply-by Denotes a time and/or date expression which indicates a guideline on

the latest time by which the sending agent would like a reply.

:protocol Introduces an identifier which denotes the protocol which the sending

agent is employing. The protocol serves to give additional context for

the interpretation of the message.

:conversation-

id

Introduces an expression which is used to identify an ongoing

sequence of communicative acts which together form a conversation. A

Conversation may be used by an agent to manage its communication

strategies and activities.

Table 1-1: FIPA ACL Message Parameters [14].

The ACL language has been used in JADE, a Java agent based platform developed by

Telecom-Italia.

1.3.2 Knowledge Query and Manipulation Language (KQML)

Knowledge Query and Manipulation Language (KQML) is older than the FIPA ACL

language. The Knowledge Query and Manipulation Language (KQML) is a protocol for

exchanging information and knowledge. The beauty of KQML language is that, all

information for understanding the content of the message is included in the message

pattern itself [6]. The basic protocol for the KQML language is defined by the following

structure,


11

(KQML-performative

: sender <word>

: receiver <word>

: language <word>

: ontology <word>

: content <expression>

…)

The KQML protocol for communications among both agents and application program is

illustrated in the following figure,

Most of the today’s agent based platforms are migrating to FIPA ACL language for the

agent communications.

1.4 Problem Statement

In this thesis an agent based platform is developed for a distribution system. This Agent

platform will serve as an automatic fault locator and reconfiguration entity for the

distribution system. The study is based on a proto-type circuit, the Circuit of the Future

(CoF) developed by Southern California Edison (SCE).

Figure 1-3: KQML communication protocol [6].


12

An agent is located at each node, load, source and switch in the distribution system.

When a fault occurs in the network, agents will start communicating to each other and

come up with the fault location. Once the fault is detected depending upon the fault, the

reconfiguration algorithm comes up with reconfigured network in which all the loads will

get supplied. The final step in the reconfiguration is to do the voltage control by using

shunt compensation and priority based load shedding.

A few assumptions made in this thesis work are listed below,

• The substation bus is modeled as an infinite bus. Assuming it is connected to the

power-grid, where it can pull up the required power.

• DG is modeled as a backup source.

• When a fault is identified, the protection is applied first and isolates the faulty lines,

before reconfiguration

1.5 Approach

The original CoF, is simplified by lumping certain loads without affecting the switch

locations and maintaining the original topology of the system. The circuit is modeled using

Distributed Engineering Workstation (DEW), power distribution software developed by

Electrical Distribution Design Inc. Power World Simulator is used for single phase

simulation of circuit of the future network. The software used to design the MAS is Java

Agent Development (JADE) framework and MATLAB. Fault detection algorithm is

developed by JADE and reconfiguration algorithm is implemented in MATLAB. This

software will be presented in detail in Chapter 3.

The agent frame work in JADE will first detect the fault location and the global agent,

which has the capability of performing load flow analysis on the circuit, will use the

MATLAB reconfiguration algorithm to reconfigure the circuit. The one solution with less

power losses and good voltage profile will be chosen as the final solution, when multiple

solutions are available for the reconfiguration. In the Reconfiguration algorithm the voltage

control is achieved by shunt compensation and priority based load shedding.


13

1.6 Outline

The outline of the remaining chapters is given in this section.

Chapter 2 will describe the literature overview of different multi-agent applications in

power distribution systems. It also covers a few other applications of multi-agents outside

the power engineering field.

A detailed description of the software packages used in this thesis is given in Chapter 3.

The advantages of this software and the suitability to the relevant applications are also

discussed.

A mathematical model, based on graph theory, for fault detection and reconfiguration

algorithms will be presented in Chapter 4. This chapter will include the modeling of the

system and the Circuit of the Future that is used in this work to perform the simulations.

Chapter 5 will present the simulations and results of the model. Finally, conclusion of this

study and the future work is detailed in Chapter 6.

14

Chapter 2

LITERATURE REVIEW

This chapter deals with comprehensive overview of the Multi-agent applications in power

system field and many other engineering disciplines. The chapter starts with the

explanation of Multi-agent system technology and its overview. Applications of Multi-agent

systems in power systems, such as in automatic fault detection, reconfiguration,

protection coordination, voltage control, reactive power control and power markets are

presented in detail manner in the later section of the chapter. Finally the chapter

concludes with explaining the Multi-agent system advantages over the traditional schemes

available and technical difficulties of Multi-agent systems.

2.1 Multi-agent systems

A Multi-agent system is a computational system in which several agents cooperate and

coordinate with each other to achieve some objectives. Multi-agent systems have the

capability to solve problems which are difficult or sometimes impossible to solve by an

individual agent. Multi-agents systems are a branch of Distribute Artificial Intelligence

(DAI). As a social community the agents are able to evolve, self produce, learn,

cooperate, and adapt. According to [6] the Multi-agent system environment has the

following characteristics,

Chapter 2: Literature Review

15

• Multi-agent environment provide an infrastructure specifying communication and

interaction protocols.

• Multi-agent environments are highly decentralized structures.

• Multi-agent environments contain agents that are autonomous and distributed, and

may be self-interested and cooperative.

Research on Multi-agent systems is still an ongoing process. Although Multi-agent

systems have shown good results, these technologies are yet to make a significant impact

in the power system area. The authors in [15] have analyzed agent based technologies

under the Strengths Weaknesses Opportunities Threat (SWOT) framework. Based on the

SWOT analysis they have envisaged the future application of agents in the power system

area. The following table shows the SWOT matrix developed by them

Strengths Opportunities • No single point of failure

• Decentralized data processing

-Fast operation

-Task distribution

• Ease of tackling complex problems

• Natural solution

• Leads to scalable, modular and

flexible solutions

• Technological advances

• Collaboration/cooperation

• Develop generic agents/MAS

• Existing hardware or software

infrastructure.

• Learn from other distributed

technologies.

Weaknesses Threats • Can lead to unstable operation

• Complexity

• Lack of understanding

• Lack of experts

• Still its in infancy

• Lack of practical application in power

area

• Unimagined problems

• Acceptability are not taken

seriously

• Centralized schemes

• Security issues

Table 2-1: SWOT matrix for current Agent Applications to power systems [15].


16

Z. Zhang et al. in [16] identified and described three major problems of today’s power

system industry. They are distributed computing, communications, and data integration.

They have suggested that the Multi-agent system technology would be a best solution for

the above problems in power systems because of the power system’s distributed nature.

They illustrated that the MAS can offer modular, flexible, and integrated approach for

many problems in the power system area. They have proposed a four stage (Analysis,

Design, Implementation, and Deployment) problem solving mechanism using Multi-agent

system.

2.2 MAS in Automatic Fault Detection and Reconfiguration

Fault detection is a very important thing in providing survivability, reliability, availability,

and efficiency of the power system. Many authors have contributed in the field of

automatic fault reconfiguration and restoration of the power system.

Authors L. Liu et al. in [17] have proposed a new agent paradigm for fault detection and

prognosis in an electric war ship. They have described the inability of traditional diagnosis

schemes over the new software agent technologies for fault detection. Prognosis is the

important thing in their work, where they can predict the future conditions in a time ahead

and can plan for the future actions to avoid them. The proposed agent architecture has

two types of agents named diagnosis agents and prognosis agents. A test case on

permanent magnet synchronous motor (PMSM) has been demonstrated and the results

were promising.

In [18] and [19], the authors have described the automatic fault detection and fault

diagnosis (FDD) mechanism using model based methods. These methods are based,

e.g., on parameter estimation, parity equations or state observers. The main aim of their

work is to create several symptoms to identify the difference between normal and faulty

systems. The fault in the system is detected by the use of classification or interference

methods. Authors in [20] and [21] proposed the fault diagnosis and detection mechanism

using analytical redundancy. In [22] K. Huang et al. had given a brief survey about the

agent technologies available for fault detection and restoration in Ship board Power


17

System (SPS). They presented a decentralized scheme of agent technology for automatic

fault detection. Several test cases have been demonstrated to prove the proposed work.

In [23], the authors presented different online techniques for fault detection.

Several authors have presented new methods of fault reconfiguration schemes since a

decade. New technologies like Agent technologies are getting more emphasis in the field

of fault detection and restoration of power systems. Many authors described the

advantages of decentralized schemes over the centralized schemes.

In, [24] the authors proposed MAS for power distribution system. Unlike the traditional

centralized reconfiguration schemes, they proposed a completely new decentralized

reconfiguration scheme. In their Multi-agent systems each agent represents a major

component in the power system such as generators, breakers, and loads. Each agent will

communicate to its neighboring agent only, which makes the whole agent system more

decentralized. The agent platform is created in Java Agent Development Framework

(JADE), and the real power system is implemented in a real time digital simulator. The

agents are implemented in iPAQs, iPAQ is a Pocket PC (PPC) developed by Hewlett

Packard. Power system is modeled in Real Time Digital Simulator (RTDS). The series

ports available in RTDS are used for the communication between RTDS and iPAQs.

However, an interface named Field Programmable Array Gate (FPGA) is developed for

the communication because of the voltage difference between RTDS and iPAQs.

In [25], the authors proposed a Multi-agent system approach for reconfiguration and

restoration of Electric warship. The proposed scheme is a decentralized one, which has

advantage of avoiding the single point failure as of in centralized reconfiguration scheme.

In their work they have introduced three kinds of agents, named Switch Agents (SAs),

Substation Breaker Agents (SBAs) and Tie Breaker Agents (TBAs). Fault detection in the

system is based on monitoring the fault currents of neighboring agents. The ship board

power system is modeled using three phase unbalanced PQ load models. Ship board

system is developed in Virtual Test Bed (VTB) and the agent platform is developed in

Matlab.

In [26], the same authors came up with a better agent model for the power system

reconfiguration and restoration problem. The Multi-agent platform is developed in Java


18

Agent Development Framework (JADE) and the power system test case is developed in

Virtual Test Bed (VTB). In their work they proposed three new agents named Switch

Agents (SAs), Load Agents (LAs), and Generator Agents (GAs). Here the agents

communicate only to their neighboring agents and act locally, making the restoration

decentralized. The restoration algorithm has certain objectives and constraints like, limit

on generation, priority of loads, and transfer capacity of lines. The three phase

unbalanced load flow is carried out by forward and backward propagation.

Takeshi Nagata et al. [27] have greatly contributed in the area of MAS for the restoration

of distribution systems. They have proposed a restoration method by joint usage of Expert

Systems (ES) and a Mathematical Programming (MP) approach. The main purpose of

their work is to determine the optimal target system for restoration using Mathematical

Programming (MP). They introduced a new topic called ‘’restorative operation cost’’, which

has decreased the number of rules in the knowledge based Expert Systems (ESs). The

Expert systems available at that time were inefficient in finding the optimal target

configuration. The authors took up the same knowledge based Expert Systems and used

Mathematical Programming methods to improve their capability of finding the optimal

target configuration. Several test scenarios are presented in their work to demonstrate the

combined technology of Expert Systems and Mathematical Programming.

The same authors in [28], [29] have proposed a Multi-agent architecture for power system

restoration problem. This architecture was fully decentralized one, in which the agents

communicate and negotiate only with their neighboring agents. In this work authors have

used two types of agents named Bus Agents (BAGs) and Facilitator Agent (FAG). BAG is

a bus agent situated at each and every bus in the network whereas, FAG is a Facilitator

Agent situated on top of all BAGs for the decision making purpose. BAGs communicate

with their neighboring agents to find out the suboptimal target configurations where as,

FAG acts as a manager for decision process. Knowledge Query and Manipulation

Language (KQML) has been used as an agent communication language.

In [30], authors Takeshi Nagata et al. proposed new agent architecture for power

distribution system. In this publication the proposed method has a new agent called

Junction Agent (JAG). JAG is a Junction Agent situated at each and every multi terminal

transmission line to monitor and control it. The authors have used Foundation for


19

Intelligent Physical Agents (FIPA) compliant XML/ACL, which is ACL (Agent

Communication Language), mounted on XML (Extensible Markup Language) for

communication between the agents. Several test scenarios are presented in their work

demonstrate the proposed algorithm for the reconfiguration and they sound promising for

the large scale power system restoration.

Later the same authors in [31] have proposed a decentralized Multi-agent architecture for

the bulk power system restoration. In this work they have introduced hierarchy among the

agents. They classified the agents into upper and lower level agents. The proposed

system consists of several Local-area Management Agents (LMAs) and Remote-area

Management Agents (RMAs). LMAs and RMAs are located in the upper level for

monitoring and control purposes. Several Load Agents (LAGs) and Generator agents

(GAGs) are situated in the lower level of physical power system network. LMAs are

responsible for the local area restoration by means of communicating and negotiating with

the neighboring LMAs. RMAs facilitate the restoration of loads within the remote area.

LAGs and GAGs correspond to load management system and generator management

system which store their respective loads and generators local information. In addition to

their local communication, they also interact with the upper level agents.

In [32], authors have proposed a more complicated four layered agent architecture for the

bulk power system restoration. A single Independent System Operator Agent (ISOAG)

which controls the timing of the simulation process forms the higher level in the agent

architecture. The second level consist several Local Area Management Agents (LAMs) for

the local area management for the restoration. Third level of architecture has several Load

Facilitator Agents (LFAGs), Generator Facilitator Agents (GFAGs), and Remote Facilitator

Agents (RFAGs). These agents will facilitate with the agents situated in the fourth layer of

architecture which are, Load Agents (LAGs) and Generator agents (GAGs). The authors

have applied the proposed MAS to bulk power system and the simulation results are very

promising.

Authors in [33] have also contributed in the field of Multi-agent system for bulk power

system restoration. The method employed in this work makes use of both centralized and

decentralized schemes to obtain the restoration solution. The agent architecture has two

hierarchical agents called, Management Agents (MGAGs) and Agents representing


20

practical power system components (PCAGs). Each PCAG monitors several agents

related to a subsystem named Generator Agents (GAGs), Substation Agent (SAGs), and

Load Agents (LAGs). One MGAG is generated to in charge of the subsystem and

provides negotiations and communications between PCAGs and with other subsystems.

2.3 MAS in Distribution System with DER

The power grid has been changing its physical infrastructure, control and communication

infrastructures because of the deregulation in the power industry. Many distributed

generation firms with less pollution and cheap operating costs came into the field of

generation, making the energy market competitive. Distributed Energy Resources (DERs)

can improve the reliability, stability, and quality of the electric grid. However, distributed

energy resources bring several problems related to connection to the electric grid and the

protection issues.

In [34], authors have proposed a hybrid Multi-agent system to achieve scalability for

control of a large network of power generation, transmission, load, and compensation

sources. Several ancillary agents are developed for system stability and harmonic and

reactive current compensation in the paper. Scalability is the greatest issue present in

huge networks such as power network, where the power network has thousands to tens of

thousands nodes. In decentralized architecture agents makes use of local information to

achieve their goals. The information could be values of voltages, currents sometimes may

be the power values. Authors have presented a hybrid agent architecture where agents

can connect not only to their parents and children but also to their siblings. Peer agents

can communicate and negotiate with each other. Peer agents will dynamically select a

leader to establish the real connection with their parents. In the proposed agent

architecture, authors have developed several compensator agents to take care of power

electronic compensators in the power system. The power electronic compensator has

several responsibilities such as reactive power generation, power flow control, harmonic

compensation, voltage regulation or dynamic control over the frequency and voltage. A

large set of decision criteria is necessary for the agents such that they can make control

decisions that will ultimately improve the power quality and reliability of the electric grid.


21

T. Hiyama et al. in [35] presented a Multi-agent based operation and control of distribution

system with dispersed power sources such as a photo-voltaic unit, a wind generation unit,

a diesel generation unit, and an Energy Capacitor System (ECS). Several simulation

studies have been performed to prove that the proposed methods have increased the

control performance even in the existence of communication delays.

In [36] the same authors proposed a reconfiguration technique using the Multi-agent

technology for the distribution system which has several kinds of distributed generation.

The distributed generation includes Diesel Generator (DG), wing generator, photo-voltaic

and ECS. ECS is an energy storage device and plays a main role in load following

operation. Computer networks are used as communication channels for communicating

the information about DGs and ECSs. Authors have introduced a coordination scheme

between diesel unit and ECS with the help of multi agents. Reconfiguration has been

done both for radial and mesh networks.

2.4 MAS in protection coordination

Many in the research community have conducted research concerning to the MAS

application for the protection coordination in power system. It is very important task for

agents in the protection coordination in order to ensure power system safety and

reliability. Selectivity, Celerity, Sensitivity, Reliability in relaying protection have already

been improved, but the system is lack of cooperation and coordination to solve the

problems. Relays play a vital role in power system operations with very low tolerance for

failure. Hence, MAS technology can be a promising solution for all of these protection

coordination problems in power network.

In [37], the authors have proposed Multi-agent system approach for the post-fault

disturbance diagnostics. Protection engineers use data from different monitoring devices

to perform the post-fault disturbance diagnostics. In the past, several heterogeneous

intelligent systems have been developed to interpret the data received and to assist the

engineers in post-fault disturbance diagnostics. Majority of the system were standalone


22

because of the system integration issues. In this paper a novel MAS architecture has

been developed named protection engineering diagnostic agents (PEDAs). The new

proposed method delivers flexibility and scalability features. These agents will integrate

Supervisory Control and Data Acquisition (SCADA) systems with new systems for Digital

Fault Recorder (DFR) for enhancing the fault record retrieval from remote DFRs. The

paper elaborates the benefits of a MAS approach and design and implementation of

PEDA.

Jinai Zhang et al. in [38] came up with new Multi-agent system architecture for relaying

protection in power system. The proposed scheme has a coordination layer where all the

agents, which are monitoring the protection in the power network, are coordinating for

protection and control of power grid independently and collectively. Authors have used the

PEDA technology, discussed in [37], to remove failures correctly and quickly. They have

mentioned several advantages of MAS over traditional methods, such as enhancing

intelligence, flexibility, reliability of the power system. In [39], the authors have proposed

an agent based methodology for current differential relay to use with a utility intranet.

Agents, as proposed in this paper are software pieces capable of searching for

information in the networks, interacting with pieces of equipment and performing tasks on

behalf of their owners (relays). Results have illustrated that the proposed scheme was

very promising agent-based differential method acting within a communication structure

i.e., intranet.

In [40] the authors have presented an agent based power protection system using

supervisors. In this paper the framework of supervisory control of discrete event system

(DESs) is applied for agent-based power protection system for achieving high reliability

and selectivity. Reliability is a measure of the degree that the protection system will

perform correctly when required, and the selectivity is an ability to recognize the fault and

to trip the circuit breaker to isolate that fault. The proposed system uses the feedback

information of events to achieve its objective goals. A detailed design procedure of a

supervisor that coordinates the behavior of relay agents to isolate fault areas through the

minimum operations of circuit breakers is presented in this paper.

A Multi-agent architecture is developed in [41] for the protection coordination of power

system having distributed generation in it. Protection of power system, which includes


23

DGs is going to be more complicated than simple radial network protection. The proposed

scheme is capable of such type of protection, High Impedance Fault (HIF) detection, fault

location and load shedding for the DG system. The complete design and implementation

of relay agents is presented in this paper. The relay agents will coordinate and

communicate with other relay agents for performing tasks of protection with autonomy.

EMTP simulation results prove that the proposed scheme is very promising MAS for

distributed generation.

The authors of [42] have presented a Multi-agent approach for protection coordination

with distributed generation in industrial power distribution system. In the proposed system

communication will play a major role besides relay settings. The agent architecture is

developed using Java Agent Development Framework (JADE). In this work the author

proposed three types of agents named relay agent, equipment agent and distributed

generation agent. The relay agent is responsible for searching the relevant information by

communication, detecting malfunctions of relay, breaker failures and DG connection

status. The equipment agent collects the local power system information and acts on the

local power system. The distributed agent will communicate with the relay agent to

provide the connection status of its own for protection coordination. In [43], the same

authors have proposed MAS architecture for substation protection with distributed

generation. In this architecture each substation is viewed as one JDE container. Each

JADE container again has number of agents in it. In this work author introduced new

agent called Substation Management Agent (SMAG). The results presented in the paper

are very promising.

The same work on protection coordination is investigated in [44], [45] and [46]. Finally,

MAS proves to be a prominent solution for today’s protection coordination issues in power

system.


24

2.5 MAS in Voltage Control

Many authors have contributed in the field of voltage control and reactive power control of

power system. In this section some of their work is illustrated. Power quality refers to two

quantities which are voltage and frequency. Frequency is the measure of balance

between the generation and load, and maintained always at nominal value for the power

system reliability. Voltage is second quantity which is always kept under certain limits and

is the major concern in the power distribution system. MAS technology is a very promising

solution for reactive power management and voltage control in power distribution system

because of agents supreme attributes such as autonomy, cooperation, Intelligence,

adaptation and social behavior.

In [47], the authors have proposed a decentralized agent mechanism for the secondary

voltage control in power-system contingencies. The secondary voltage control mainly

applied for the generator Automatic Voltage Regulators (AVRs) to improve the power

system voltage stability. In this paper, different types of voltage controllers, such as an

Automatic Voltage Regulator (AVR), a Static Var Compensator (SVC) or Static

Synchronous Compensator (STATCOM), is treated as an agent. The agents which are

electrically close are responsible for their own area voltage control. When an agent

senses voltage violation it activates its reactive power reserve. If the voltage violation is

not cleared by its own reactive power reserve then the agent starts asking reactive power

from its neighbor agents. The agents with sufficient reactive power will respond to the

respective agent for voltage support. The agents share their common goal and achieve

goal by communication and coordination. The proposed scheme is applied on simple

power system and the results are promising.

The authors in [48] have presented optimal coordination work for Multi-agent based

secondary voltage control in power system. The agent architecture has a set of execution

agents and a coordination agent. Individual voltage controllers, such as an AVR, a SVC

and a STATCOM represent an execution agent. Under normal conditions, Multi-agent

system based voltage co-ordination works as a conventional secondary voltage control

and supports Var/Voltage control. When system runs in to contingencies, the Multi-agent

system uses contract net protocol to realize coordination and cooperation among voltage

control agents for eliminating voltage violation. During the contingency situations the


25

execution agents (EA) will act independently and try to restore the voltage in their areas. If

the execution agents fail in restoring the voltage within limits, they send message to

coordination agent (CA). The coordination agent then will coordinate with its execution

agents to restore the voltage profile. Authors have used New England 39-bus system as a

test system to demonstrate the proposed agent architecture and the results were up to the

mark.

2.6 MAS in Power Markets

Power Market is the other major area in the power system, where Multi-agent systems are

exploited to a wide range. In [49], the authors have proposed a Multi-agent system for

power markets, which makes use of game theory principles. They have developed

specialized intelligent software agents that perform negotiations on behalf of their human

counterparts, and then suggest the market strategies that the human can adopt. The

game theory principles are used for negotiation protocol. The agent model required not to

exchange any sort of trustworthy important information between coalition partners. This is

the fact where game theory differs from the proposed work in this paper. In [50], the same

authors applied this work to a three player market model. This particular model has three

generation companies where each individual company wants to maximize its profits.

Hyungna Oh et al. in [51] have investigated about, why price spikes occur in deregulated

wholesale market and how efficiently they can be mitigated. The authors have presented

MAS architecture to replicate a spot market. The proposed agent architecture has six

autonomous adaptive agents representing six generating stations and one Independent

System Operator agent representing an Independent System Operator (ISO). The firm

learns about the market and its competitor’s behavior by comparing actual market

outcomes with predicted outcomes based on an estimate of their own residual demand

curve. The estimated residual demand curve is updated each interval using a kalman

filter. In [52], the same authors presented a agent based work to test how spot prices

affected by forward contracts.


26

Leigh Testfation et al. in [53] developed a Java agent based whole sale power market

simulator. The simulator is referred as Agent-based Modeling of Electricity Systems

(AMES) and is designed in accordance with core Federal Energy Regulatory

Commission’s (FERC)’s recommended design features and operating over a realistically

rendered transmission grid. This framework has several learning mechanisms through

which generation companies can maximize their profits and utilization factors. Authors

have taken a 5-bus test case to illustrate the simulator features.

Several other papers such as [54], [55], and [56] also contributed for Multi-agent system

application in power markets.

2.7 MAS in other engineering disciplines

Distributed Artificial Intelligence (DAI) is very promising and developing technology in

many fields of engineering. Many engineering applications which needs monitor and

control aspects are opting for Multi-agent system technology, which is the part of DAI.

Several firms in engineering today are exploiting the Multi-agent technology to harness its

strengths towards technical problem solving. A few of them are discussed in brief below.

Highway Traffic Management is one area where researchers are using agent based

technologies for making them effective and reliable. The traffic congestion in peak hours

results in loss of productivity by conveyors, waste of valuable time, increase in accidents,

and a risk of public health with the increasing pollution from the vehicles. In [57], the

authors have proposed a multi agent system for traffic management which can mitigate

the traffic congestion in Automated Highway System (AHS). Autonomous agents

cooperate and coordinate with each other to solve problem, such as conflicts, incidents

and synchronization. The behavior of agent is described by colored petri nets.

Due to ever increasing intensity of air traffic and its safety issues, MAS proves to be

promising solution for Air Traffic Control (ATC). In [58] authors have proposed a agent

architecture for ATC which employs several agents playing roles of assistant aircraft

crews and assistant air traffic control operator of approach zone. ATC operator agent


27

(ATCO) is responsible for coordination of airliner’s entries in to the ATC zone. Its

knowledge base (KB) consists of rule set representing temporal constraints on concurrent

usage of approach schemes. Airliner agent is responsible for control over airline trajectory

along assigned arrival scheme. The authors developed a software prototype intended for

validation of the Multi-agent model of ATC.

Mine detection is the other specific area where distributed artificial Intelligence techniques

such as MAS are widely used. In [59], the authors have studied and implemented hybrid

architecture to implement mine detection, obstacle avoidance and route planning by using

MAS. There are different techniques to design the mine exploration. The authors have

followed frontier based exploration. In this type of exploration the agent builds its own map

and starts searching on it. After the coordination mechanism post exploration route

planning is carried out. This will enable the agent to generate safe route plans for the

given target destinations.

There are several other fields in which MAS technologies have been used for variety of

reasons. For example, computer games, logistics, robotics, graphics and e-commerce are

some of the fields where extensive research is going on for embedding the Multi-agent

systems.

2.8 Limitations and Technical challenges of MAS

It is recently that Multi-agent technologies are shifting from laboratory to utility areas to

harness its potential features. As stated earlier in the chapter, Multi-agent systems are

used in many power systems and many other engineering areas.

Despite the advantages of Multi-agent systems and its growing awareness, some basic

questions often arise from the researchers and industrialists, in particular, industrial

partners when discussing Multi-agent systems and their role in power engineering. The

basic questions are as follows,


28

• What are the benefits offered by MAS over traditional technologies?

• What are the major differences between MAS and traditional technologies?

• To which sort of problems MAS can be applied?

These are the key questions regarding to the Multi-agent systems and often confusing.

Many researchers have struggled answering these questions. D. J. McArthur et al. in [60]

were able to answer the above questions. MAS can encapsulate a particular task or set of

functionalities, and acts in a similar way to object oriented programming. In fact MAS is

adding another level of abstraction to object oriented programming, not only the internal

data structures are hidden but also action of agents (methods) can be hidden. This

particular fact makes MAS pioneer in energy markets, where certain things need to be

kept in secret between agents. According to [60], the main advantages of MAS are as

follows,

• Agent autonomy and agent encapsulation

• Open Multi-agent system architecture.

• Can be used as a platform for distributed systems

• Fault tolerance

The key difference between MAS and its counterparts such as AI techniques, Expert

systems, Model based reasoning (MBR) systems, and Artificial Neural Networks (ANNs)

is the notion of autonomy and interaction protocols available in MAS. MAS offer a great

amount of autonomy feature to its agents, where agents can independently decide on

actions to achieve their goals. Another great feature of MAS is that, it has a richer set of

interaction protocols which supports negotiations. The authors in [60] have explained

clearly that, this MAS technology can be applied to any engineering discipline which

requires a decentralized problem solving mechanism. In spite of many advantages of

MAS, there are certain technical challenges that are yet to be overcome to allow most

effective implementation of multi-agent systems in the power engineering community. The

technical challenges are listed below


29

• Platforms

Today in the market several Multi-agent platforms are available. A judicious

selection of agent platform is required to ensure the long-term compatibility and

required robustness for online applications.

• Toolkits

Reuse of existing agent behaviors and capabilities is an important task. Thus,

selection of toolkits to perform these activities is a major thing for the benefit of the

whole community.

• Intelligent agent design

Standardized agent design architecture is needed for educating the new

researchers and industrial implementers about MAS.

• Agent communication languages and ontologies

• Specific data standards

• Security issues

• Mobile agents

Many researchers are interested in moving or mobile agents. These agents

actually move from one point to another point in the network. As of now no credible

reason for using this approach is apparent.

These above listed things are few technical challenges which have to be overcome such

that MAS technologies are widely used in power engineering area. Beyond above

technical challenges and implementation issues, the lack of experience in the use of multi-

agent system technology in the industry is an obvious concern of both utilities and

manufacturers. In [61], the same authors stressed on design and implementation issues of

Multi-agent systems.

The other main challenge would be interfacing the power system software with the agent

platform. Normally the agent platforms are created in some Java based systems, where

as power system models are created using traditional power system software. Many times

there exists a big difficulty in interfacing these two platforms. In [62], the authors proposed

a methodology to interface the MAS application, JADE to the power system application,

which is developed in Virtual Test Bed (VTB). The authors have realized this interface

through a communication mechanism called Common Object Request Broker Architecture

(CORBA).

30

Chapter 3

SIMULATION SOFTWARE

This chapter briefly describes about various software packages used in this thesis work

for modeling the agent-based power distribution system. It is clear from the previous

chapter that MAS greatly depend on software simulations and it is very important to

choose suitable software packages for the agent based simulations.

Among several technical challenges listed in the previous chapter, selecting a appropriate

software package for simulating MAS is the major technical challenge. Several options are

analyzed for simulating both power distribution system and Multi Agent System. Finally,

MATLAB, MATPOWER, Power World Simulator, Distributed Engineering Workstation

(DEW), and Java Agent Development Framework (JADE) are chosen to simulate the

power distribution system and agent model in this thesis work.

The whole Multi-agent system model interfaced to power distribution system is depicted in

the following figure. The bottom layer in the figure represents physical network of power

distribution system. Agents are situated at all components of power system such as,

buses, loads, distributed generators, etc. These agents form upper layer in the

architecture. There is a great need of an interfacing tool for interfacing the distribution

system model to the agent architecture.

Chapter 3: Simulation Software

31

3.1 MATLAB ®

MATLAB ®, Matrix Laboratory, is a high level technical language and numerical

computing environment for algorithm development, data visualization, data analysis, and

numeric computation. MATLAB solves technical oriented tasks faster than traditional

languages such as C, C++, and FORTRAN [49]. MATLAB was first adopted by control

design engineers, later it is been used in many other scientific applications including

image processing, communications, etc. Some of the key features of MATLAB are listed

as followed [63],

• Developing environment for managing code, files, and data

• High level language for technical computing.

• Tools for creating custom graphical user interfaces.

• Functions for integrating MATLAB based methods to external applications, such as

C, C++, FORTRAN, Java, and Microsoft Excel.

• 2-D and 3-D graphics functions for visualizing data.

Agent Layer

Power System Layer

Interface

Figure 3-1: Architecture of the Simulation Model.


32

MATLAB is more suitable for engineering applications because of its operations based on

matrix and vectors. MATLAB is faster in engineering tasks, because of the basic tasks like

variable declaration, memory allocation, and data type specifications are default. MATLAB

uses Object Oriented Programming (OOP) thus, adds the object oriented features to it.

This feature enables to develop complex technical computing applications easily. When

using well defined classes, object-oriented programming can significantly increase the

code reuse and, make program easier to maintain and extend. Besides the OOP

attributes, MATLAB has Graphical User Interface (GUI) tool to create and visualize

customized interfaces.

In this thesis work MATLAB object-oriented programming is used to create the

reconfiguration algorithms and the plotting functions. In addition to that, a graphical user

interface for the reconfiguration has been developed using MTALAB GUI.

3.2 MATPOWER

MATPOWER is a MATLAB based power system simulation package. It is a collection of

MATLAB m-files to solve the load flow and optimal power flow problems of power system

[64]. It is indeed a simulation package for researchers and educators, where one can

easily make modifications to the original code according to one’s application. Ray D.

Zimmerman is the main contributor for this power simulation software. The need for

MATLAB based power flow and optimal power flow born out of the computational

requirements of the PowerWeb project at Cornell University [65]. MATPOWER is an open

source software and freely downloadable at [64]. MATPOWER is used for load flow

calculations of Circuit of the Future (CoF) distribution network in this thesis work.

3.3 Java Agent Development Framework

JADE (Java Agent Development Framework) is a software package for developing agent-

based applications in compliance with Foundation for Intelligent Physical Agents (FIPA)

specifications for interoperable and intelligent multi-agent systems [66]. JADE is a middle-

ware application completely developed using Java language. JADE has extensive


33

graphical tools for the debugging and deployment phases. The key advantage offered by

JADE is its Graphical User Interface (GUI), through which one can visualize the

development phases and agent communications.

Today, JADE is the most popular agent middle-ware that implements an agent platform

and a development framework. JADE was developed in 1998 by the Research and

Development department of Telecom Italia, in 2000. JADE is released as open source

software under the license of Lesser General Public License (LGPL). JADE software is

freely downloadable from [66] with source code, documentation, and some wealthy

information.

3.3.1 JADE Architecture

JADE architecture has several agent containers that can be distributed over the network.

Agents live in the containers are the Java processes which provide the JADE run-time

and, all the services needed for instantiating and executing the agents [67]. In every JADE

framework, there is a special container, called the main container. It is the first container

to be launched and all other possible containers must join to a main container by

registering with it. By default main container is named ‘Main Container’ and the remaining

containers are named ‘Container-1’, ‘Container-2’, etc. The main duties of main container

are listed below [67],

• Creating and Managing the container table (CT), which is the registry of the object

references and transport addresses.

• Creating and Managing the Global Agent Descriptor Table (GADT), which is the

registry of all agents present in the JADE platform including their current status

and location.

• Hosting the AMS agent, which provides agent management and white page

service.

• Hosting the DF agent, this provides default yellow page service.


34

Figure 3-2: Relationship between main architectural elements [67].

The two main agents named Agent Management System (AMS) and Directory Facilitator

(DF) are instantiated when the main container is launched.

• The Agent Management System (AMS)

Agent Management System is the special agent which supervises and monitors

the whole platform. It is the main hub for all agents present in the container in

order to access the white pages of the platform and to get Agent Identifier (AID).

All agents life cycle are managed by AMS. The registering process of all agents

with AMS is done automatically by JADE at start-up.

• The Directory Facilitator (DF)

Directory Facilitator (DF) agent implements the yellow age service. These services

are used by any agent wishing register its services or search for other available

services. Multiple DFs can start concurrently in order to distribute the yellow page

services across several domains.


35

Besides AMS and DF agents, the one more important service according to FIPA is a

Message Transport Service (MTS). An MTS manages all message exchanges within and

between platforms. It supports several Message Transport Protocols (MTPs) to promote

interoperability between different non-JADE platforms [67]. There are several other agent

tools such as, Dummy Agent, Sniffer Agent, and Introspector Agent to facilitate the

developing and debugging of agent platforms.

• The Dummy Agent

The dummy agent is a very simple agent which is used to simulate other agent’s

behavior by sending the custom Agent Communication Language (ACL)

messages. When an application agent has been launched, a Dummy agent can be

used to stimulate it by sending user-specified messages and analyzing its reaction

in terms of received messages.

• The Sniffer Agent

This is a very useful agent tool in terms of debugging the project. This agent tool

enables to visualize conversations between agents. The conversations between

agents are shown by arrows; each arrow contains the message content. In the

Sniffer agent GUI, different agent communications can be marked with different

colors.

• The Introspector Agent

The Introspector agent is useful to debug the behavior of a single agent. This

agent will control and monitors the agent sent and received messages, scheduled

behaviors of agent, and the agent life cycle. In simple words, this agent is

responsible for introspection of agent execution.

The following figure shows a typical JADE GUI for the remote management, monitoring

and controlling of the agents present in platform.


36

Figure 3-3: JADE Graphical User Interface (GUI).

The following figure shows the output of JADE run-time environment, when JADE is launched.

Figure 3-4: JADE run-time environment.


37

3.3.2 JADE Programming

JADE programming is similar to the standard JAVA programming. The creation of an

agent is done by a simple class jade.core.Agent and implementing the setup() method.

The setup() method intended to include agent initializations. The actual job performed by

the agent is defined within the ‘behaviors’ of agent. The ‘behavior’ property tells the agent

about its typical operations. The typical operations, an agent performs in its setup()

method are: showing a GUI, opening the connection to database, registering the services

to provide the yellow pages, and starting the initial behavior [67].

According to FIPA specifications each agent is identified by a unique ‘agent identifier’ in

Java Agent Development Framework. In JADE, the getAID() method is used to retrieve

the local agent identifier. The AID object includes a globally unique name (GUID) and a

number of addresses. These AIDs are very important entities in agent communication.

As discussed earlier agent communication is the most important feature of Multi-agent

systems. Each agent that is created has a mailbox where the messages are passed in a

queue when it receives messages from other agents. The messages are compliant with

FIPA standards. The ACL messages are sent to one agent from another agent by using

send() method. The receiving agents can receive the message by calling receive()

method. Agent will be alive even after completing all its responsibilities. doDelete()

method is used in order to terminate it.

JADE has been used in this thesis work for developing the agent platform for the Circuit of

the Future (CoF) distribution system.

3.4 Distributed Engineering Workstation© (DEW)

Distributed Engineering Workstation is a distribution system software and an open

architecture environment with integrated data and applications, analysis, design, and

operation modules access database data and exchange data through common functions.

The distribution engineering tool is designed to meet the analysis, planning, design, and

operation needs of today’s distribution systems. DEW offers myriad tools such as, power


38

flow, load estimation, line impedance calculations, fault analysis, capacitor placement,

phase balancing and many more [68]. Some of the add-ons of this software can be bought

separately from the vendor. The open architecture allows external application modules to

be added to the workstation.

Executive, Database, Graphical User Interface (GUI), Application Programmer Interface

(API), and Application modules are the major components of the Distributed Engineering

Workstation (DEW) [69]. These major functional parts of the DEW are depicted in the

following figure.

The main operations of the workstation are managed by Executive. It provides interface

between API, GUI, and database modules. The key feature of API is to integrate the

application modules into the framework without disturbing other parts of the workstation.

All the applications utilize the GUI functions for user interaction and display. The standard

data schema available in DEW makes data accessible and shareable by all applications.

The data from Supervisory Control and Data Acquisition System (SCADA) such as,

financial, statistical, geographical, operations data are utilized for the application through

an interface called, Database Access Integration Service (DAIS).

SCADA

DEW

Distribution Database

Corporate Data

DEW Executive GUI

Application Modules

API Data Access

DAIS

User

DAIS

Figure 3-5: Organizational structure of DEW [69].


39

The following figure shows the Graphical User Interface (GUI) of DEW. The right side

pane of the DEW main page has different power system entities. These entities can be

dragged into the DEW workspace to build the one-line model of the distribution system. In

this thesis work DEW has been used to model the Circuit of the Future (CoF) distribution

network.

Figure 3-6: Distributed Engineering Workstation’s Working Environment.

3.5 Power World Simulator

Power World simulator is an interactive power system simulation package designed to

simulate high voltage power system operation. The software contains a highly effective

power flow analysis package capable of efficiently solving systems with up to 100,000

buses [70]. Some of the key tools offered by Power World Simulator are listed below,


40

• Load Flow Analysis

• Optimal Power Flow (Linear programming method)

• Security Constrained Optimal Power Flow (SCOPF)

• Available Transfer Capabilities (ATC)

• PVQV curve tool

• Transmission line Parameter Calculator

• Contingency Analysis

• Fault Analysis

This software is a very good tool for modeling the power system. The key feature of this

software is its graphical user interface (GUI), where one needs to drag and drop the

power system elements to build up one’s own case. Different tools available in Power

World Simulator are listed above. The following figure shows the snapshot of one line

diagram of Circuit of the Future (CoF) created in Power World Simulator.

Figure 3-7: Snapshot of Circuit of the Future one line diagram


41

Power World Simulator has a very good facility of visualizing the load flow results. Every

time we run the load flow all the results will be stored and viewed through a GUI entity

called Model Explorer. The model explorer has a explore pane, which contains a list of the

objects contained in the power system model. The entries in the explore pain which are

grayed out indicates that the elements are not defined in the model. The explore pain

again divided into eight sub folders named, Recent, Network, Aggregation, Solution

Details, Case Information and Auxiliary, Contingency Analysis, Optimal Power Flow, and

Transient Stability. The following figure shows a snapshot of Power World Simulator’s

Model Explorer.

Figure 3-8: Snapshot of Model Explorer of Power World Simulator [70].

42

Chapter 4

MATHEMATICAL MODEL AND

ALGORITHMS FOR MAS

This chapter mainly explains the mathematical model used behind the Multi-agent

framework in this thesis. Multi-agent architecture has its own mathematical model to serve

as its foot-print. Once the mathematical model is build then it can be translated into

software packages for simulation purpose. This chapter presents a set of new algorithms

for fault detection and reconfiguration. The fault detection algorithm is based on Graph

theory, and the reconfiguration algorithm is based on load-flows.

4.1 Mathematical Model

4.1.1 Graph Theory

Graph theory is the study of graphs used extensively in the fields of mathematics and

computer science. A ‘graph’ is defined as, collection of vertices or ‘nodes’ and collection of

edges that connects pairs of vertices [71]. The graphs are primarily categorized as either

directed graphs or undirected graphs. The major difference between above two graphs is,

the direction between two vertices. Graph theory has several applications such as network

Chapter 4: Mathematical Model and Algorithms for MAS

43

analysis, study of molecules in physics and chemistry, and social network analysis in

sociology. The main application of graph theory lies in modeling and analyzing traffic

networks. Power system is a vast network consisting tens of thousands of vertices, can be

modeled using graph theory.

4.1.2 Graph Theory representation of Power System

The power system network is modeled as a graph G as shown in the Figure 4-1. It has a

single parent node. The power distribution feeders form the spanning tree T. Each node in

graph G represents power system elements such as loads, sources and buses. These

nodes are connected by edges. The edges in spanning tree T represent transmission

lines connecting every pair of buses. The edges in G, but not in T represent switches.

These switches are used for restoration purposes during the fault and left open during

normal healthy conditions of power grid. In the proposed MAS, agents are place at each

and every node and switch of the power distribution network. The following table

describes some of the key notations used in the mathematical model.

Notation Description

G Graph, represents the power network.

T Spanning tree, represents radial feeders.

N Set of nodes in the graph.

E Set of directed edges in the graph, represent power lines.

s Power source, represents substation or a distributed generator.

S(G) Set of edges in G, but not in T, represents set of switches in the system

Table 4-1: Description of the notations used in Mathematical model.


44

In the above graph model s represents the substation. Each branch 321 nnns →→→ ,

654 nnns →→→ , 987 nnns →→→ forms the spanning tree. The edges in the graph,

but not in the trees are switches, S1-S6. In the spanning tree every node except the root

node has in-degree1 1. That is, each node n in T can be reached by a directed path in T

form the source s to n.

For each edge )(TEe∈ , let

=)(eS { SXX ⊆: , such that XeT U)( − is a connected graph}

1 In-degree of a vertex in a graph is defined as the number of edges coming into the vertex.

s n4 n5 n6

n1 n2 n3

n7 n8 n9

S1 S3S2

S6S5S4

Figure 4-1: Graph model of power distribution system.


45

4.2 Fault Detection Algorithm

The following section describes about the Fault Detection algorithm used in this thesis

work. Fault detection is the first thing in restoring the faulty power network. Major input to

this algorithm is power flowing into each load.

Let us assume that the fault location detected by the node agents is at fn . Suppose a

node agent n identifies that the power flowing into its node is zero, then on the unique

directed path from s to n , say knnnns →→→→→ ......321 , where kk nnn ,= will

request 1−kn , whether it has power. Subsequently, 1−kn will request 2−kn for its power.

When for some i , in does not have power, but 1−in has power, the fault edge can be

identified as, ),( 1 iif nne −= [2].

Once the fault location is identified, the edge fe is isolated and agents that are controlling

each switch )( feSX ∈ will communicate and coordinate with each other to decide on the

particular switch or switches that have to be switched on, so that XeT f U)( − will

continue to be a fully supplied network [2].

4.3 Fault Reconfiguration & Restoration Algorithm

The reconfiguration algorithm employed in this thesis work is completely based on load-

flows. This algorithm is embedded inside the Global agent for reconfiguration of the faulty

network. Depending on the number of faults in the system, there are three different

algorithms to reconfigure the power distribution network. They are,

• Reconfiguration algorithm for single fault.

• Reconfiguration algorithm for double fault.

• Reconfiguration algorithm for three faults.


46

4.3.1 Reconfiguration algorithm for single fault

The fault location, fe , has been found from the previous algorithm and possible switches

are explored through this algorithm for reconfiguration. This algorithm searches for all the

possible combinations of switches, which has to operate in order to reconfigure the circuit.

If there is voltage violation in the circuit, the algorithm employs shunt compensation and

load shedding methods to improve the voltage profile. When there are multiple solutions

available for reconfiguration, the best solution is chosen in regards of voltage profile and

less power losses. Each time the algorithm executes a scenario, the solution is stored in a

database. Next time when the same scenario happens the reconfiguration would run from

the database rather than running through the algorithm. The algorithm is described

through the flowchart in the Figure 4-2.

4.3.2 Reconfiguration algorithm for two faults

The solution from single fault reconfiguration algorithm is used in this algorithm. When we

use solution form above algorithm, some times the load flow will not converge. In this case

this algorithm itself searches for additional switches to reconfigure the faulty circuit. If the

reconfigured circuit has voltage violations, then the algorithm employs shunt

compensation and priority based load shedding for voltage profile improvement. Similar to

above algorithm this algorithm also has a database to store the solutions of the scenarios.

When there are multiple solutions, the best solution in regards to voltage profile and less

power losses is chosen.

4.3.3 Reconfiguration algorithm for three faults

This algorithm is also similar to the above two fault reconfiguration algorithm. The solution

from the single fault reconfiguration algorithm is used in this algorithm. The voltage control

is done in the same manner using shunt compensation and priority load shedding.


47

Isolate the faulty line.

Start

Search for one more switch to operate Si i=1 to 6.

Switch on the shunt compensation

Print the results and plot the network Write the solution to database.

End

Yes

Switch on each switch at a time. Si i=1 to 7.

If load flow converging

Run the load flow

0.9<|V|<1.1

No

No

0.9<|V|<1.1

Yes

No

0.9<|V|<1.1

Shed the loads priority wise

0.9<|V|<1.1No

Yes

Yes

Yes

No

If multiple solutions, select

the less loss configuration

If scenario exists in database

No

Run the solution form database

Yes

Figure 4-2: Flow Chart for Fault Reconfiguration


48

4.3.4 Objective functions and Constraints

The main objective functions of the reconfiguration algorithm are

• Isolating the faulty part of the network and restoring the remaining part of the

network with good voltage profile.

• Always supplying the critical loads.

These objective functions are subjected to following constraints.

• Power flow across the transmission line should be kept under its thermal limits.

• Voltage magnitudes of all buses should be kept under the limits of 0.9<|V|<1.1.

• Less loss configuration is chosen for the reconfiguration when multiple solutions

exist.

The reconfiguration topology achieved will comply with the objective functions and the

constraints listed above. There is a major consideration for voltage profile of the

reconfigured network. The voltage control in the algorithm is achieved by means of shunt

compensation. When the shunt compensation is not sufficient to meet the voltage limits, it

starts shedding the non priority loads. Algorithm employs a database technique to store all

the scenarios such that, when the same scenario happens the reconfiguration runs form

database rather than through algorithm. This technique greatly reduces the algorithm

execution time, which is the major concern in algorithm efficiency.


49

4.4 Illustration

In this section a particular scenario is illustrated for better understanding of the

reconfiguration algorithm. This illustration is based on simplified CoF circuit as shown in

Figure 4-3. In this figure, iN are the nodes, iL are the loads, iS are the switches, s is the

substation and DG is the Distributed Generator modeled as a backup source. Let us

assume that the fault is in )( 21NNe f = .

The load agents in L1, L2 and L3 will detect that there is no power flowing into their

respective loads. Hence, these three load agents will communicate with their respective

neighbors in the incoming tree from 1234 NNNN →→→ for their flows. When all these

nodes detect that they don’t have power flow, the MAS can locate the fault in N1N2. The

load agents will report this fault location to the global agent, which has the potential to

reconfigure the circuit. This agent has the reconfiguration algorithm discussed in section

4.3. The algorithm execution is described in following steps.

• Isolates the faulty link, i.e. link 1-2.

• There are three switches which can individually reconfigure the network. They

are, 1S , 2S and 3S .

• By switching on the above switches individually, the voltage limits are not met.

• Now the algorithm searches for one more switch to operate by keeping the 1S

on, which can reconfigure the network with no voltage violation. Same procedure

repeats with 2S and 3S .

• Finally, algorithm finds two solutions which have no voltage violations. They are

{ 2S , 5S } and { 1S , 5S }.

• Out of this two solutions, { 1S , 5S } is chosen because of the less losses.

• Finally the results are printed and the reconfigured network is plotted.


50

Figure 4-3: Representation of Simplified CoF.

As discussed above, the algorithm comes up with the best possible switch combination

{ 1S , 5S } for the reconfiguration, when there is a fault at link 1-2. The same steps are

explained elaborately in the section 5.4.1 in the next chapter.

51

Chapter 5

SIMULATION AND RESULTS

In this chapter, the fault detection algorithm and the reconfiguration algorithm discussed in

the previous chapter are tested on simplified CoF distribution system. The fault detection

algorithm is developed in JADE, where as the reconfiguration algorithm is developed in

MATLAB. Power World Simulator and DEW have been used for modeling and designing

the simplified CoF.

5.1 Proposed MAS Architecture

The proposed MAS architecture for fault detection and reconfiguration is illustrated in

Figure 5-1. According to the hierarchy, there are mainly two categories of agents available

in the proposed MAS architecture. They are,

• Local Agents

• Global Agent.

Load Agents (LAGs) and Switch Agents (SAGs) fall into the category of local agents.

These agents represent loads and switches in the power distribution system respectively.

Any load agent can be associated with a number of switch agents. Whenever there is a

fault in the system, load agents and switch agents start communicating and coordinating

each other in order to locate the fault location in the network. Once the fault is detected,

the fault location is reported to the Global Agent (GAG) by load agents. The

Chapter 5: Simulation and Results

52

reconfiguration algorithm is embedded inside the GAG, such that whenever it receives

fault location it starts reconfiguring the network to reroute the power to critical loads.

The local agents such as LAGs and SAGs only can communicate to its neighboring

agents. These agents will have local information pertaining to that particular region only.

Local agents communicate and coordinate in the sense of decentralized manner, where

as the Global agent operates much in centralized manner. Because of this fact, the

proposed MAS architecture is a hybrid MAS architecture. In this hybrid architecture the

fault detection is done in a decentralized manner and the fault reconfiguration is done in a

centralized fashion. The main responsibilities of each agent can be summarized as

below,

• Load Agents (LAGs)

These agents will continuously monitor the loads present in the system. Whenever

they sense no power, they start communicating their power flow information to

their neighboring agents. Finally they will come up with the fault location based on

fault detection algorithm. Once the fault is detected, the fault location is reported to

Global agent (GAG).

LAG – Load Agent; SAG – Switch Agent; GAG – Global Agent Communication

LAG LAG LAG LAG

SAG

SAG

SAG

GAG GAG

Figure 5-1: Proposed Multi-agent system architecture.


53

• Switch Agents (SAGs)

These agents are the dummy agents, which keep status of the switch position and

coordinate with the GAG and LAGs during fault detection and reconfiguration.

• Global Agent (GAG)

The Global agent is very important entity in fault reconfiguration. It has all the

information regarding network and has the capability of doing load flows. The

reconfiguration algorithm is embedded inside the GAG. GAG will call its Voltage

Control routine, whenever there is a need for voltage control. Every time GAG

runs the scenario it stores it in its database for learning. When the same scenario

happens in the future, GAG recalls its experience and reconfigures the network

without using the fault reconfiguration algorithm. This is the main step, GAG

employs in reducing the execution time of fault reconfiguration. The GAG and its

functionalities are described in Figure 5-2.

Fault Reconfiguration

Voltage Control

Database

GAG

Figure 5-2: Global Agent (GAG) functionalities.


54

5.2 Circuit of the Future (CoF)

The fault detection and reconfiguration algorithms are tested on the proto-type distribution

system, the Circuit of the Future (CoF). This is the distribution system modeled and

designed by Southern California Edison (SCE). The same distribution system is modeled

using Power World Simulator and DEW. The modeled circuit using DEW is shown in

Figure 5-3.

The CoF has two sources, one is the substation and the other is a distributed generator

(DG). The substation is connected with three outgoing feeders, which will serve the loads

in the distribution system. This system has 7 switches/disconnectors for the purpose of

reconfiguration of faulty network. The distribution network has 14 loads demanding 26.69

MW of real power and 14.578 MVar of reactive power. The network has 14 shunt

capacitor banks for the purpose of voltage control. It has two DGs, one for real power

generation and other for reactive power generation.

For simplicity purposes the original circuit has been modified by lumping all the loads

together without affecting the basic topology and key features of the original CoF circuit.

This distribution network from now on will be termed as Simplified CoF. This Simplified

CoF has been modeled in Power World Simulator, and in shown in Figure 5-4. The key

components of Simplified CoF are listed below,

• 18 Buses

• 17 Distribution lines

• 7 Switchable Transmission lines (Switches/disconnectors)

• 11 fixed loads ( P: 26.69 MW, Q:14.578 MVAR, modeled as ‘PQ buses’)

• 1 DG (P: 3 MW, modeled as ‘PV bus’)

• 1 Substation (Modeled as ‘Slack Bus’)

• 7 shunt capacitor banks (4.5 MVAR)


55

Figure 5-3: The Circuit of the Future modeled in DEW [2].


56

The 11 loads, which exist in the Simplified COF, have load priorities depending upon the

importance of particular load. Loads such as Hospital loads and Industrial loads are given

higher priority than any residential loads in the system. This distinction of the loads is

necessary in the load shedding process for voltage control. The loads with lower priority

will get shed before the higher priority loads. There are three higher priority loads (L3, L8

and L10), three lowest priority loads (L7, L9 and L11) and five middle priority loads (L1, L2, L4

L5 and L6) in the network.

Figure 5-4: The Simplified CoF modeled in Power World Simulator.


57

5.3 Fault Detection and Reconfiguration Algorithm

As discussed earlier fault detection is done by Load Agents (LAGs) and reconfiguration is

achieved by Global agent (GAG). LAGs will follow the fault detection algorithm explained

in section 4.2, for fault detection. There are two main ways in which we can instantiate the

agents.

• Instantiating JADE agents from reading the power flow report, which is generated

by Power World Simulator.

• Instantiating JADE agents by passing necessary command line arguments to the

JADE platform.

Both cases are efficient and correct as per the results are concerned. The only difference

between them is, generating the command line arguments needed to instantiate the JADE

agents.

The first step towards the simulation process is to simulate the fault in Power World

Simulator. Once the fault is simulated, the corresponding load flow report is sent to

Microsoft Excel. This excel file is then read by JADE function for instantiating the JADE

agents. Once the LAGs, SAGs and GAG are created in the JADE platform, the fault is

detected by LAGs and SAGs. The restoration needed for isolating the faulty part of the

network and supplying the remaining part of the network is achieved by Global agent

(GAG). The whole process is explained through the following flowchart,


58

Yes

Read the load flow result Excel File from JADE

If flow of any LAG = 0

No

Find the In-line Neighboring LAG

Run the Load flow in Power World

Simulator

Send Request for flow Message to Neighboring

LAG

Receive Reply

If Reply Content = 0

Yes

No

Fault Location Identified

Kill the active LAGs

No Fault

Start

End

Report the Fault location to GAG to reconfigure the circuit

Figure 5-5: Flow chart for Fault Detection and Reconfiguration.


59

5.3.1 Test Case 1: For a Single Fault

Let us test the system for a single fault in line N2N3 in Feeder 1. First of all, the fault is

simulated in Power World Simulator and the corresponding network is depicted in Figure

5-6.

Figure 5-6: Simplified CoF with fault in line 2-3.

Once the fault is simulated in Power World Simulator, It produces following load flow

report which is saved in excel file as shown below in Figure 5-7. This report has the power

flow values of all loads. Because of the line fault at line 2-3, the loads L2 and L3 are not

supplied at all. The same information can be observed from the following excel sheet,

where MW values of ID: 2 and ID: 3 are zero.


60

Figure 5-7: Power World Simulator’s Load flow report for line fault at line 2-3.

When JADE MAS is run, the load agents in the faulty feeder will only get created. These

agents will communicate and coordinate with each other to find out the fault location. The

message passing between agents is clearly shown in JADE sniffer agent GUI.

Figure 5-8: JADE message passing between agents for line fault at line 2-3.


61

All the LAGs will be created in the particular feeder, but only those agents which sense

zero power start communicating with their incoming neighboring load agents regarding

whether they have power or not. In this particular scenario, the load agents LAG2 and

LAG3 has no power. Once they sense zero power, they start sending the request

message to their neighboring load agents, to LAG1 and LAG2 respectively. Here LAG1

will report its power flow value to LAG2, and LAG2 will report to LAG3. According to the

fault detection algorithm explained in section 4.2, the fault is detected at line 2-3, i.e.

between nodes N2N3. This solution can be found on Java output console, as shown in

Figure 5-9.

Figure 5-9: Java output console showing results of Test Case 1.

This fault location is now reported to Global Agent, which calls the MATLAB

reconfiguration algorithm to reconfigure the network. The MATLAB reconfiguration

algorithm is explained in section 4.3. Once the algorithm finds the reconfigured network, it

plots the network and prints the results. The reconfigured network for fault at line 2-3 is

shown in following Figure 5-10.


62

Subs

tatio

n

11.

000

Bus

2

0.95

8

Bus

3

0.93

8

Bus

4

0.92

4

Bus

13

0.99

3

G 2

4.13

7| 1

0.16

7

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.95

6L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

0

G

3.00

0| 6

.664

Bus

6

0.95

4

Bus

7

0.95

4

Bus

8

0.94

1

Bus

9

0.92

1

Bus

15

0.98

5

Bus

17

0.97

6

Bus

18

0.96

8

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.91

9

Bus

11

0.91

6

Bus

12

0.91

3

L 1

.260

| 0.

680

L 1

.330

| 0.

781

03-J

an-2

009

19:0

2:02

Tota

l gen

erat

ion:

27.

137|

16.

831

Tota

l load

: 26

.690

| 14.

578

Tota

l shu

nt

:

0.00

0| 0

.000

Tota

l loss

es

: 0

.447

| 2.

254

SW1

SW2

SW3

SW4

SW5

SW6

SW7

Figure 5-10: MATLAB plot depicting reconfiguration network for fault at line 2-3.


63

5.3.2 Test Case 2: For Multiple Faults

Let us assume that there are two faults in Feeder 1 and Feeder 2 at N1N2 and N1N5. The

procedure would be as same as in the previous test case. This scenario is been simulated

first in Power World Simulator and the respective command line arguments are sent to

JADE platform to instantiate the agents. The message passing between the Load Agents

(LAGs) have been observed and shown in following figure.

Figure 5-11: JADE message passing between agents for line faults at line 1-2 and 1-5.

The fault is detected according to the fault detection algorithm and shown on output Java

console as shown below,

Figure 5-12: Java output console showing results of Test Case 2.

Once the faults are detected by LAGs, the information is then sent to GAG for

reconfiguration. The reconfiguration network is shown in Figure 5-13.


64

Figure 5-13: MATLAB plot depicting reconfiguration network for fault at line 1-2 and 1-5.

Subs

tatio

n

11.

000

Bus

2

0.92

3

Bus

3

0.91

5

Bus

4

0.90

0

Bus

13

0.98

9

G 2

4.77

7| 3

.638

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.93

2L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

0

G

3.00

0| 1

3.67

9

Bus

6

0.93

0

Bus

7

0.92

9

Bus

8

0.92

9

Bus

9

0.93

5

Bus

15

0.96

6

Bus

17

0.95

6

Bus

18

0.94

8

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.93

9

Bus

11

0.95

0

Bus

12

0.96

1

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

15:5

2:26

Tota

l gen

erat

ion:

27.

777|

17.

317

Tota

l load

: 26

.690

| 14.

578

Tota

l shu

nt

:

0.00

0| 0

.000

Tota

l loss

es

: 1

.087

| 2.

740

Subs

tatio

n

11.

000

Bus

2

0.96

9

Bus

3

0.96

0

Bus

4

0.94

6

Bus

13

0.98

9

G 2

4.86

0| 1

.711

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.91

1L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

0

G

3.00

0| 1

5.19

2

Bus

6

0.91

1

Bus

7

0.91

1

Bus

8

0.91

2

Bus

9

0.92

3

Bus

15

0.96

2

Bus

17

0.95

2

Bus

18

0.94

4

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.92

8

Bus

11

0.94

2

Bus

12

0.95

6

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

15:5

2:27

Tota

l gen

erat

ion:

27.

860|

16.

903

Tota

l load

: 26

.690

| 14.

578

Tota

l shu

nt

:

0.00

0| 0

.000

Tota

l loss

es

: 1

.170

| 2.

325

Subs

tatio

n

11.

000

Bus

2

0.91

7

Bus

3

0.91

1

Bus

4

0.90

8

Bus

13

0.98

4

G 2

6.42

8| -6

.156

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.91

8L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

2

G

3.00

0| 8

.334

Bus

6

0.91

8

Bus

7

0.91

8

Bus

8

0.92

0

Bus

9

0.91

8

Bus

15

0.95

2

Bus

17

0.94

7

Bus

18

0.94

1

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.92

1

Bus

11

0.93

8

Bus

12

0.95

3

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

15:5

2:28

Tota

l gen

erat

ion:

29.

428|

2.1

78To

tal lo

ad

:

26.6

90| 1

4.57

8To

tal s

hunt

:

0.

000|

0.0

00To

tal lo

sses

:

2.7

38|-1

2.39

9

Subs

tatio

n

11.

000

Bus

2

0.91

5

Bus

3

0.91

1

Bus

4

0.90

3

Bus

13

0.99

6

G

9.70

8| -3

.466

L 0

.000

| 0.

000

L 0

.000

| 0.

000

L 2

.560

| 1.

380

S 0

.000

| 1.

000

L 0

.000

| 0.

000

S 0

.000

| 0.

500

Bus

5

0.91

5L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

1

G

3.00

0| 7

.370

S 0

.000

| 0.

500

Bus

6

0.91

6

Bus

7

0.91

6

Bus

8

0.91

8

Bus

9

0.93

2

Bus

15

0.96

9

Bus

17

0.97

2

Bus

18

0.97

5

S 0

.000

| 0.

500

L 2

.070

| 1.

118

L 2

.420

| 1.

306

S 0

.000

| 0.

500

L 2

.670

| 1.

442

L 0

.000

| 0.

000

S 0

.000

| 1.

000

Bus

10

0.93

6

Bus

11

0.95

1

Bus

12

0.96

5

L 0

.000

| 0.

000

S 0

.000

| 0.

500

L 1

.330

| 0.

781

04-J

an-2

009

15:5

2:28

Tota

l gen

erat

ion:

12.

708|

3.9

04To

tal lo

ad

:

11.9

80|

6.52

9To

tal s

hunt

:

0.

000|

4.5

00To

tal lo

sses

:

0.7

28|

1.87

5

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7


65

5.4 MATLAB Reconfiguration

This section is mainly concentrated on MATLAB reconfiguration which is used by Global

Agent (GAG) for reconfiguration of faulty network. Many key features of the fault

reconfiguration algorithm are explained with several scenarios. A Graphical User Interface

(GUI) is developed for easy user interaction with the fault reconfiguration algorithm and to

carry out the simulations. The GUI is shown in the following figure.

Figure 5-14: MATLAB GUI for Fault Reconfiguration.


66

The basic functionalities of the above GUI are explained below,

• BASIC CKT: This push button will display the basic network configuration of

Simplified CoF.

• Line Fault: The user can specify up to three faults and create the faults in the

system.

• CLEAR DATA: Clears the data in the Line fault section.

• CLEAR: Clears the MATLAB command prompt.

• Single Fault Reconfiguration: Reconfigures the circuit based on fault location

specified in Line Fault 1 section.

• Double Fault Reconfiguration: Reconfigures the circuit based on fault locations

specified in Line Fault 1 and Line Fault 2 sections.

• Three Fault Reconfiguration: Reconfigures the circuit based on fault locations

specified in Line Fault 1, Line Fault 2 and Line Fault 3.

The basic Simplified CoF network is shown in the Figure 5-15. In this plot each and every

bus is marked with its bus number and its voltage magnitude. The loads are marked with

their real and reactive power consumptions. In the same way generators are marked with

read and reactive power generations. The switches are represented with red rectangular

boxes when they are open and turn in to green when they are closed. In the bottom left

corner of the plot, information related to total generation, total load and total losses in the

system are shown. Any transmission line which is in normal operation is shown with solid

line, where as transmission line which is opened, is shown with dashed line.


67

Subs

tatio

n

11.

000

Bus

2

0.96

9

Bus

3

0.96

0

Bus

4

0.94

6

Bus

13

0.99

3

G 2

4.15

1| 1

0.05

6

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.94

0L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

0

G

3.00

0| 6

.664

Bus

6

0.93

8

Bus

7

0.93

7

Bus

8

0.92

9

Bus

9

0.91

0

Bus

15

0.98

5

Bus

17

0.97

6

Bus

18

0.96

8

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.90

8

Bus

11

0.90

4

Bus

12

0.90

2

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

15:5

7:23

Tota

l gen

erat

ion:

27.

151|

16.

720

Tota

l load

: 26

.690

| 14.

578

Tota

l shu

nt

:

0.00

0| 0

.000

Tota

l loss

es

: 0

.461

| 2.

142

SW1

SW2

SW3

SW4

SW5

SW6

SW7

Figure 5-15: MATLAB plot depicting basic Simplified CoF.


68

5.4.1 Test Case 3: For a Line Fault

Let us assume that the fault is at line 1-2. In order to simulate this fault, one has to enter

the ‘From bus’ and ‘To bus’ as 1 and 2 in the reconfiguration GUI line fault section. Once

these fields are entered the reconfiguration can be achieved using fault reconfiguration

algorithm as explained in the section 4.3. The obtained reconfiguration circuit would

satisfy all the objectives and constraints specified in section 4.3.4.

The reconfiguration algorithm execution key steps are explained as below,

1. When there is a fault at line 1-2, the possible single switch combinations for

reconfiguring the network according to fault reconfiguration algorithm would be,

},,{)( 321 SSSeS f =

The algorithm first tries closing the switches 1S , 2S , and, 3S individually, and the voltage

profiles for each combination would be shown as below,

2 4 6 8 10 12 14 16 180.6

0.7

0.8

0.9

1

1.1

Bus numbers

Bus

Vol

atge

mag

nitu

de

Figure 5-16: MATLAB plot depicting voltage profile with S1 closed for line fault at 1-2.


69

2 4 6 8 10 12 14 16 180.6

0.7

0.8

0.9

1

1.1

Bus numbers

Bus

Vol

atge

mag

nitu

de


2 4 6 8 10 12 14 16 18

0.7

0.8

0.9

1

1.1

Bus numbers

Bus

Vol

tage

mag

nitu

des



70

2. From the above figures Figure 5-16, Figure 5-17 and Figure 5-18, one can observe

that none of the individual switch configuration gives the good voltage profile, i.e. voltage

magnitude with ±10% tolerance. Now the algorithm searches for one more switch to

operate by keeping the 1S on, which can reconfigure the network with no voltage

violation. The same procedure repeats with 2S and 3S .

3. Finally, the fault reconfiguration algorithm finds two switch combinations which have no

voltage violations, they are { 2S , 5S } and { 1S , 5S }. The following figures show the voltage

profile of the reconfiguration network with above switch configurations.

2 4 6 8 10 12 14 16 180.6

0.7

0.8

0.9

1

1.1

Bus numbers

Bus

Vol

atge

mag

nitu

des



71

2 4 6 8 10 12 14 16 180.6

0.7

0.8

0.9

1

1.1

Bus numbers

Bus

Vol

tage

mag

nitu

des


4. Hence there are two solutions for reconfiguring the line fault at 1-2. As explained in the

section 4.3.4, whenever there are multiple solutions available for reconfiguration the

algorithm chooses the configuration with less power loss. The reconfigured network with

{ 1S , 5S } has 1.087 MW real power loss where as the reconfigured network with { 2S , 5S }

has 1.151 MW of real power loss. Since, the reconfiguration with { 1S , 5S } closed will yield

less power loss, this configuration is selected as the final solution for reconfiguration. The

final reconfigured network for line fault at 1-2 is shown in the following figure.


72


Subs

tatio

n

11.

000

Bus

2

0.92

3

Bus

3

0.91

5

Bus

4

0.90

0

Bus

13

0.98

9

G 2

4.77

7| 3

.638

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.93

2L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

0

G

3.00

0| 1

3.67

9

Bus

6

0.93

0

Bus

7

0.92

9

Bus

8

0.92

9

Bus

9

0.93

5

Bus

15

0.96

6

Bus

17

0.95

6

Bus

18

0.94

8

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.93

9

Bus

11

0.95

0

Bus

12

0.96

1

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

17:4

0:39

Tota

l gen

erat

ion:

27.

777|

17.

317

Tota

l load

: 26

.690

| 14.

578

Tota

l shu

nt

:

0.00

0| 0

.000

Tota

l loss

es

: 1

.087

| 2.

740

SW1

SW2

SW3

SW4

SW5

SW6

SW7


73

5.4.2 Test Case 4: For two Line Faults

Let us assume that there are two faults at lines 1-2 and 14-15. The main aim of this

scenario is to explain the voltage control part in Fault Reconfiguration algorithm. As

explained before in the section 5.4.1, the algorithm tries searching for the best switch

configuration in regards of voltage profile and power loss. But, in certain cases it is difficult

to achieve the good voltage profile with only switch selection. In these types of cases the

Fault Reconfiguration algorithm uses its voltage control routine to achieve good voltage

profile for the reconfigured network.

In this scenario, the algorithm comes up with switch combination { 1S , 5S , 7S } , for

reconfiguring the network. The voltage profile with this switch combination is shown in the

following figure,

2 4 6 8 10 12 14 16 180.6

0.7

0.8

0.9

1

1.1

Bus numbers

Bus

Vol

tage

mag

nitu

de

Figure 5-22: Plot depicting voltage profile with S1, S5 & S7 closed for line faults at 1-2 & 14-15, without voltage control.


74

From the above figure it is clear that there is voltage violation in the reconfigured network.

Now, the algorithm calls its voltage control routine to build a good voltage profile. This

voltage control routine first tries putting all the shunt compensation available in the

network. If the voltage violation still exists, it starts priority based shedding of the loads. In

this scenario, when the shunt compensation of 4.5 MVAR is provided in the reconfigured

circuit, the voltage violation disappears. The fact can be witnessed from following figure.

2 4 6 8 10 12 14 16 180.6

0.7

0.8

0.9

1

1.1

Bus numbers

Bus

Vol

tage

mag

nitu

des

Figure 5-23: Plot depicting voltage profile with S1, S5 & S7 closed for line faults at 1-2 & 14-15 with

shunt compensation.

The voltage profile in the above Figure 5-23 has no voltage violation. The shunt

compensation of 4.5 MVAR helped in boosting up the bus voltages. The final reconfigured

network is shown in following Figure 5-24.


75

Figure 5-24: MATLAB plot depicting reconfiguration network for fault at lines 1-2 and 14-15.

Subs

tatio

n

11.

000

Bus

2

0.92

3

Bus

3

0.91

5

Bus

4

0.90

0

Bus

13

0.98

9

G 2

4.77

7| 3

.638

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.93

2L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

0

G

3.00

0| 1

3.67

9

Bus

6

0.93

0

Bus

7

0.92

9

Bus

8

0.92

9

Bus

9

0.93

5

Bus

15

0.96

6

Bus

17

0.95

6

Bus

18

0.94

8

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.93

9

Bus

11

0.95

0

Bus

12

0.96

1

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

18:3

3:49

Tota

l gen

erat

ion:

27.

777|

17.

317

Tota

l load

: 26

.690

| 14.

578

Tota

l shu

nt

:

0.00

0| 0

.000

Tota

l loss

es

: 1

.087

| 2.

740

Subs

tatio

n

11.

000

Bus

2

0.96

9

Bus

3

0.96

0

Bus

4

0.94

6

Bus

13

0.99

3

G 2

4.15

1| 1

0.05

5

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.94

0L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

0.98

5

G

3.00

0| 6

.707

Bus

6

0.93

8

Bus

7

0.93

7

Bus

8

0.92

9

Bus

9

0.91

0

Bus

15

0.97

2

Bus

17

0.98

2

Bus

18

0.97

4

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.90

8

Bus

11

0.90

4

Bus

12

0.90

2

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

18:3

3:50

Tota

l gen

erat

ion:

27.

151|

16.

763

Tota

l load

: 26

.690

| 14.

578

Tota

l shu

nt

:

0.00

0| 0

.000

Tota

l loss

es

: 0

.461

| 2.

185

Subs

tatio

n

11.

000

Bus

2

0.99

0

Bus

3

0.98

6

Bus

4

0.98

5

Bus

13

0.99

1

G 2

4.52

1| -2

.684

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.99

3L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

0.98

4

G

3.00

0| -0

.152

Bus

6

0.99

2

Bus

7

0.99

1

Bus

8

0.98

9

Bus

9

0.98

3

Bus

15

0.97

9

Bus

17

0.98

6

Bus

18

0.98

2

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.98

3

Bus

11

0.98

4

Bus

12

0.98

2

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

18:3

3:51

Tota

l gen

erat

ion:

27.

521|

-2.8

36To

tal lo

ad

:

26.6

90| 1

4.57

8To

tal s

hunt

:

0.

000|

0.0

00To

tal lo

sses

:

0.8

31|-1

7.41

4

Subs

tatio

n

11.

000

Bus

2

0.92

2

Bus

3

0.91

4

Bus

4

0.90

7

Bus

13

0.99

1

G 2

4.71

2| 3

.949

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

S 0

.000

| 1.

000

L 5

.340

| 2.

988

S 0

.000

| 0.

500

Bus

5

0.93

0L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

0.97

3

G

3.00

0| 9

.614

S 0

.000

| 0.

500

Bus

6

0.92

8

Bus

7

0.92

7

Bus

8

0.92

4

Bus

9

0.92

4

Bus

15

0.94

1

Bus

17

0.96

8

Bus

18

0.96

3

S 0

.000

| 0.

500

L 2

.070

| 1.

118

L 2

.420

| 1.

306

S 0

.000

| 0.

500

L 2

.670

| 1.

442

L 2

.670

| 1.

442

S 0

.000

| 1.

000

Bus

10

0.92

5

Bus

11

0.93

2

Bus

12

0.93

9

L 1

.260

| 0.

680

S 0

.000

| 0.

500

L 1

.330

| 0.

781

04-J

an-2

009

18:3

3:51

Tota

l gen

erat

ion:

27.

712|

13.

562

Tota

l load

: 26

.690

| 14.

578

Tota

l shu

nt

:

0.00

0| 4

.500

Tota

l loss

es

: 1

.022

| 3.

485

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7


76

5.4.3 Test Case 5: For three Line Faults

Let us assume that there are three line faults at 1-2, 14-15, and 11-12. In this scenario,

the algorithm comes up with switch combination { 1S , 5S , 7S } , for reconfiguring the

network. The voltage profile with this switch combination is shown in the following figure,

2 4 6 8 10 12 14 16 180.6

0.7

0.8

0.9

1

1.1

Bus numbers

Bus

Vol

atge

mag

nitu

des

Figure 5-25: Plot depicting voltage profile with S1, S5 & S7 closed for line faults at 1-2, 14-15

& 11-12, without voltage control.

From the above figure it is clear that there is voltage violation. In this scenario, the

algorithm calls the voltage control routine to achieve the good voltage profile. This time

not only the shunt compensation but also the priority based load shedding is done for

providing good voltage profile to the reconfigured network. The low priority loads such as,

L11, L9, L7 & L1 are shed and shunt compensation of 4.5 MVAR are provided in order to

achieve good voltage profile. The final reconfigured network is shown in the following

figure.


77

Subs

tatio

n

11.

000

Bus

2

0.92

3

Bus

3

0.91

5

Bus

4

0.90

0

Bus

13

0.98

9

G 2

4.77

7| 3

.638

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.93

2L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

0

G

3.00

0| 1

3.67

9

Bus

6

0.93

0

Bus

7

0.92

9

Bus

8

0.92

9

Bus

9

0.93

5

Bus

15

0.96

6

Bus

17

0.95

6

Bus

18

0.94

8

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.93

9

Bus

11

0.95

0

Bus

12

0.96

1

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

19:0

5:12

Tota

l gen

erat

ion:

27.

777|

17.

317

Tota

l load

: 26

.690

| 14.

578

Tota

l shu

nt

:

0.00

0| 0

.000

Tota

l loss

es

: 1

.087

| 2.

740

Subs

tatio

n

11.

000

Bus

2

0.96

9

Bus

3

0.96

0

Bus

4

0.94

6

Bus

13

0.99

3

G 2

4.15

1| 1

0.05

5

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.94

0L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

0.98

5

G

3.00

0| 6

.707

Bus

6

0.93

8

Bus

7

0.93

7

Bus

8

0.92

9

Bus

9

0.91

0

Bus

15

0.97

2

Bus

17

0.98

2

Bus

18

0.97

4

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.90

8

Bus

11

0.90

4

Bus

12

0.90

2

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

19:0

5:13

Tota

l gen

erat

ion:

27.

151|

16.

763

Tota

l load

: 26

.690

| 14.

578

Tota

l shu

nt

:

0.00

0| 0

.000

Tota

l loss

es

: 0

.461

| 2.

185

Subs

tatio

n

11.

000

Bus

2

0.96

9

Bus

3

0.96

0

Bus

4

0.94

6

Bus

13

0.99

2

G 2

4.15

4| 8

.343

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.95

3L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

0

G

3.00

0| 7

.993

Bus

6

0.95

1

Bus

7

0.95

0

Bus

8

0.94

4

Bus

9

0.93

1

Bus

15

0.98

2

Bus

17

0.97

2

Bus

18

0.96

4

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

0.93

0

Bus

11

0.93

0

Bus

12

0.98

1

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

19:0

5:13

Tota

l gen

erat

ion:

27.

154|

16.

336

Tota

l load

: 26

.690

| 14.

578

Tota

l shu

nt

:

0.00

0| 0

.000

Tota

l loss

es

: 0

.464

| 1.

758

Subs

tatio

n

11.

000

Bus

2

1.05

0

Bus

3

1.04

7

Bus

4

1.04

8

Bus

13

0.99

4

G 2

4.38

3| -8

.318

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

1.04

9L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

0.98

8

G

3.00

0| 0

.420

Bus

6

1.04

9

Bus

7

1.04

8

Bus

8

1.05

0

Bus

9

1.06

4

Bus

15

0.97

9

Bus

17

0.98

8

Bus

18

0.98

3

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

10

1.06

7

Bus

11

1.07

0

Bus

12

0.98

0

L 1

.260

| 0.

680

L 1

.330

| 0.

781

04-J

an-2

009

19:0

5:14

Tota

l gen

erat

ion:

27.

383|

-7.8

98To

tal lo

ad

:

26.6

90| 1

4.57

8To

tal s

hunt

:

0.

000|

0.0

00To

tal lo

sses

:

0.6

93|-2

2.47

6

Subs

tatio

n

11.

000

Bus

2

0.94

3

Bus

3

0.93

5

Bus

4

0.92

8

Bus

13

1.00

0

G 1

1.43

3| 4

.011

L 0

.000

| 0.

000

L 2

.130

| 1.

150

L 2

.560

| 1.

380

S 0

.000

| 1.

000

L 0

.000

| 0.

000

S 0

.000

| 0.

500

Bus

5

0.94

8L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

0.99

1

G

3.00

0| 1

.020

S 0

.000

| 0.

500

Bus

6

0.94

6

Bus

7

0.94

5

Bus

8

0.94

1

Bus

9

0.93

8

Bus

15

0.97

4

Bus

17

0.98

9

Bus

18

0.99

2

S 0

.000

| 0.

500

L 2

.070

| 1.

118

L 2

.420

| 1.

306

S 0

.000

| 0.

500

L 2

.670

| 1.

442

L 0

.000

| 0.

000

S 0

.000

| 1.

000

Bus

10

0.93

9

Bus

11

0.93

9

Bus

12

0.97

4

L 0

.000

| 0.

000

S 0

.000

| 0.

500

L 1

.330

| 0.

781

04-J

an-2

009

19:0

5:15

Tota

l gen

erat

ion:

14.

433|

5.0

31To

tal lo

ad

:

14.1

10|

7.68

0To

tal s

hunt

:

0.

000|

4.5

00To

tal lo

sses

:

0.3

23|

1.85

2

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

Figure 5-26: MATLAB plot depicting reconfiguration network for fault at lines 1-2, 14-15 and 11-12.


78

5.4.4 Test Case 6: Special Scenario

In this section, a special scenario is discussed. Let us assume there are two faults at lines

9-10 and 10-11. In this particular scenario, there is no way that load L7 is supplied. In

these types of scenarios, the algorithm modifies its network data file in a manner to

exclude the particular load which is not supplied. Once this has been done, the algorithm

executes in a normal way to reconfigure the rest of the circuit. The MATLAB output

command window for this scenario is shown in the following Figure 5-27,

Figure 5-27: MATLAB output command window showing solution of Test Case 5.

The algorithm comes up with switch combination 5S for reconfiguring the network with

faults at 9-10 and 10-11. It can be observed from the above figure that there is no voltage

violation in the network with switch configuration 5S . The final reconfigured network for this

scenario is shown in the following Figure 5-28.


79

Subs

tatio

n

11.

000

Bus

2

0.96

9

Bus

3

0.96

0

Bus

4

0.94

6

Bus

13

0.99

2

G 2

2.84

8| 7

.345

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.96

3L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

0

G

3.00

0| 7

.993

Bus

6

0.96

2

Bus

7

0.96

1

Bus

8

0.95

7

Bus

9

0.94

8

Bus

15

0.98

2

Bus

17

0.97

2

Bus

18

0.96

4

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

11

0.98

1

Bus

12

0.98

1L

1.3

30|

0.78

104

-Jan

-200

9 19

:15:

39To

tal g

ener

atio

n: 2

5.84

8| 1

5.33

8To

tal lo

ad

:

25.4

30| 1

3.89

7To

tal s

hunt

:

0.

000|

0.0

00To

tal lo

sses

:

0.4

18|

1.44

1

Subs

tatio

n

11.

000

Bus

2

0.96

9

Bus

3

0.96

0

Bus

4

0.94

6

Bus

13

0.99

2

G 2

2.84

8| 7

.345

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

L 5

.340

| 2.

988

Bus

5

0.96

3L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

0

G

3.00

0| 7

.993

Bus

6

0.96

2

Bus

7

0.96

1

Bus

8

0.95

7

Bus

9

0.94

8

Bus

15

0.98

2

Bus

17

0.97

2

Bus

18

0.96

4

L 2

.070

| 1.

118

L 2

.420

| 1.

306

L 2

.670

| 1.

442

L 2

.670

| 1.

442

Bus

11

0.98

1

Bus

12

0.98

1L

1.3

30|

0.78

104

-Jan

-200

9 19

:15:

40To

tal g

ener

atio

n: 2

5.84

8| 1

5.33

8To

tal lo

ad

:

25.4

30| 1

3.89

7To

tal s

hunt

:

0.

000|

0.0

00To

tal lo

sses

:

0.4

18|

1.44

1

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

Figure 5-28: MATLAB plot depicting reconfiguration network for fault at lines 9-10 and 10-11.


80

5.4.5 Test Case 7: Special Scenario

In this section, a special scenario related to Test Case # 3 is discussed. Let us assume

there is a fault at distribution line 1-2. In this particular scenario, the way the

reconfiguration algorithm reconfigures the network is modified. From the previous chapter,

it is clear that the reconfiguration algorithm searches for second switch for getting voltage

support after operating the first switch. In this scenario, the reconfiguration algorithm

operates the first switch to reconfigure the network, and then uses the voltage control

technique for getting the voltage support. Here the reconfiguration algorithm doesn’t

search for the second switch for the voltage support. This scenario helps in understanding

different formulations of reconfiguration algorithm. The final reconfigured network is shown

in Figure 5-29.

The final reconfiguration plot clearly shows the difference between execution of Test Case

# 3 and Test Case # 7. In Test Case # 3, two switches S1 and S5 are operated in order to

reconfigure the network and to get the good voltage profile. In the present case, only the

switch S1 is operated and to build the voltage profile voltage control techniques such as

shunt compensation and priority based load shedding are used. In this case, total of 4.5

MVAR of shunt compensation and loads L11, L9, L7, and L1 are turned off for providing

good voltage profile across the network.

If one doesn’t want to use the voltage control techniques and want to build the voltage

profile using possible switch configurations, then one can adopt the procedure followed in

Test Case # 3. If one is reluctant about switching ON more switchable distribution lines,

then one can follow the present case. It’s purely one’s discretion to choose the way

he/she wants to reconfigure the network. Hence, this scenario just visualizes different

formulations of fault reconfiguration and their pros and cons.


81

Subs

tatio

n

11.

000

Bus

2

0.92

9

Bus

3

0.92

1

Bus

4

0.91

3

Bus

13

1.00

1

G 1

1.45

6| 5

.910

L 0

.000

| 0.

000

L 2

.130

| 1.

150

L 2

.560

| 1.

380

S 0

.000

| 1.

000

L 0

.000

| 0.

000

S 0

.000

| 0.

500

Bus

5

0.93

5L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

1

G

3.00

0| -0

.373

S 0

.000

| 0.

500

Bus

6

0.93

2

Bus

7

0.93

1

Bus

8

0.92

5

Bus

9

0.91

6

Bus

15

0.99

4

Bus

17

0.99

7

Bus

18

1.00

0

S 0

.000

| 0.

500

L 2

.070

| 1.

118

L 2

.420

| 1.

306

S 0

.000

| 0.

500

L 2

.670

| 1.

442

L 0

.000

| 0.

000

S 0

.000

| 1.

000

Bus

10

0.91

6

Bus

11

0.91

3

Bus

12

0.91

0

L 0

.000

| 0.

000

S 0

.000

| 0.

500

L 1

.330

| 0.

781

11-F

eb-2

009

11:3

4:16

Tota

l gen

erat

ion:

14.

456|

5.5

37To

tal lo

ad

:

14.1

10|

7.68

0To

tal s

hunt

:

0.

000|

4.5

00To

tal lo

sses

:

0.3

46|

2.35

7

Subs

tatio

n

11.

000

Bus

2

0.98

8

Bus

3

0.91

8

Bus

4

0.91

0

Bus

13

1.00

1

G 1

4.78

5| 7

.893

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

S 0

.000

| 1.

000

L 0

.000

| 0.

000

S 0

.000

| 0.

500

Bus

5

0.93

3L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

1

G

3.00

0| -0

.373

S 0

.000

| 0.

500

Bus

6

0.93

1

Bus

7

0.93

0

Bus

8

0.92

0

Bus

9

0.91

1

Bus

15

0.99

4

Bus

17

0.99

7

Bus

18

1.00

0

S 0

.000

| 0.

500

L 2

.070

| 1.

118

L 2

.420

| 1.

306

S 0

.000

| 0.

500

L 2

.670

| 1.

442

L 0

.000

| 0.

000

S 0

.000

| 1.

000

Bus

10

0.91

0

Bus

11

0.90

8

Bus

12

0.90

5

L 0

.000

| 0.

000

S 0

.000

| 0.

500

L 1

.330

| 0.

781

11-F

eb-2

009

11:3

6:02

Tota

l gen

erat

ion:

17.

785|

7.5

20To

tal lo

ad

:

17.4

20|

9.46

7To

tal s

hunt

:

0.

000|

4.5

00To

tal lo

sses

:

0.3

65|

2.55

3

Subs

tatio

n

11.

000

Bus

2

0.97

9

Bus

3

0.97

6

Bus

4

0.92

7

Bus

13

1.00

1

G 1

4.67

8| 7

.364

L 3

.310

| 1.

787

L 2

.130

| 1.

150

L 2

.560

| 1.

380

S 0

.000

| 1.

000

L 0

.000

| 0.

000

S 0

.000

| 0.

500

Bus

5

0.95

1L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

1

G

3.00

0| -0

.373

S 0

.000

| 0.

500

Bus

6

0.95

0

Bus

7

0.94

9

Bus

8

0.94

1

Bus

9

0.92

8

Bus

15

0.99

4

Bus

17

0.99

7

Bus

18

1.00

0

S 0

.000

| 0.

500

L 2

.070

| 1.

118

L 2

.420

| 1.

306

S 0

.000

| 0.

500

L 2

.670

| 1.

442

L 0

.000

| 0.

000

S 0

.000

| 1.

000

Bus

10

0.92

7

Bus

11

0.92

5

Bus

12

0.92

2

L 0

.000

| 0.

000

S 0

.000

| 0.

500

L 1

.330

| 0.

781

11-F

eb-2

009

11:3

6:58

Tota

l gen

erat

ion:

17.

678|

6.9

90To

tal lo

ad

:

17.4

20|

9.46

7To

tal s

hunt

:

0.

000|

4.5

00To

tal lo

sses

:

0.2

58|

2.02

3

Subs

tatio

n

11.

000

Bus

2

0.92

9

Bus

3

0.92

1

Bus

4

0.91

3

Bus

13

1.00

1

G 1

1.45

6| 5

.910

L 0

.000

| 0.

000

L 2

.130

| 1.

150

L 2

.560

| 1.

380

S 0

.000

| 1.

000

L 0

.000

| 0.

000

S 0

.000

| 0.

500

Bus

5

0.93

5L

0.9

30|

0.50

2

DG 14

1.00

0

Bus

16

1.00

1

G

3.00

0| -0

.373

S 0

.000

| 0.

500

Bus

6

0.93

2

Bus

7

0.93

1

Bus

8

0.92

5

Bus

9

0.91

6

Bus

15

0.99

4

Bus

17

0.99

7

Bus

18

1.00

0

S 0

.000

| 0.

500

L 2

.070

| 1.

118

L 2

.420

| 1.

306

S 0

.000

| 0.

500

L 2

.670

| 1.

442

L 0

.000

| 0.

000

S 0

.000

| 1.

000

Bus

10

0.91

6

Bus

11

0.91

3

Bus

12

0.91

0

L 0

.000

| 0.

000

S 0

.000

| 0.

500

L 1

.330

| 0.

781

11-F

eb-2

009

13:0

6:45

Tota

l gen

erat

ion:

14.

456|

5.5

37To

tal lo

ad

:

14.1

10|

7.68

0To

tal s

hunt

:

0.

000|

4.5

00To

tal lo

sses

:

0.3

46|

2.35

7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7

SW1

SW2

SW3

SW4

SW5

SW6

SW7



82

5.5 Algorithm Execution Duration

In this section both the Fault Detection algorithm and Fault Reconfiguration algorithm are

tested for their execution durations. Many times the efficiency of algorithm is related to its

execution time. The lesser time the algorithm takes to execute certain function, the

efficient is the algorithm. The following Figure 5-29 shows the execution time taken for

100 samples of fault detection processes. It is observed that the average time taken for

fault detection is 22.4 ms.

0 20 40 60 80 10015

20

25

30

35

40

45

50

time

(ms)

Execution time for Fault Detection

data 1 y mean

Figure 5-30: Plot of Execution Time for Fault Detection Algorithm [2].

The following Figure 5-30 shows the execution time taken for fault reconfigurations for 50

sample scenarios. The average execution time is 1.8796 sec.

10 20 30 40 500

1

2

3

4

5

6

time(

secs

)

execution timemean excution time

Figure 5-31: Plot of Execution Time for Fault Reconfiguration Algorithm.


83

As explained earlier in the chapter, the Global Agent has data base, which keeps record

of all the scenarios which happened earlier. When the same scenario happens in the

future GAG recalls its experience and reconfigures the circuit based on its experience.

Because of this fact the fault reconfiguration execution time is drastically decreased. The

same scenarios as of in Figure 5-30, are run again to test the execution times of Fault

Reconfiguration algorithm. The following plot shows the execution times for the same

scenarios as of in Figure 5-30 with accessing data base.

10 20 30 40 500.015

0.02

0.025

0.03

0.035

0.04

time(

secs

)

execution time

mean time

Figure 5-32: Plot of Execution Time for Fault Reconfiguration Algorithm accessing data base.

When GAG accesses its data base to run the reconfiguration, the execution time taken by

the algorithm is greatly reduced. It is observed that the average execution time for fault

reconfiguration is 25 ms, which sounds a very promising achievement in algorithm

development.

These execution times are only concerning to the software implementation. Actual

execution times in the real world implementation would be longer than the times specified

here. This is due to the software and hardware interface and communication delays.

84

Chapter 6

CONCLUSION AND FUTURE WORK

Multi-agent systems are part of Distributed Artificial Intelligence (DAI). For over a decade

many researchers have proposed myriad applications of MAS in different engineering

fields. Recently MAS are being explored and applied in the area of power engineering.

Power systems are becoming more complicated and the present control techniques are

insufficient for safer and reliable operation. MAS would be a perfect solution for large and

complicated systems like power systems. MAS are autonomous, intelligent, distributed

and provide faster solutions for today’s power system complex problems.

6.1 Conclusion

In this thesis a MAS model is proposed for Fault detection and Fault reconfiguration. The

proposed Multi-agent system makes use of both centralized and decentralized methods to

make up for the disadvantages of each method. The proposed model of MAS is tested on

proto-type distribution network named Circuit of the Future (CoF). The network has three

outgoing distribution feeders which supplies number of loads in the distribution network.

CoF has number of switchable transmission lines (switches) for the purpose of re-routing

the power during fault reconfiguration and restoration of the network.

In the proposed MAS model, there are two hierarchical agents named Local Agents and

Global agent. The operation of Local Agents is in decentralized manner where as the

Chapter 6: Conclusion and Future Work

85

operation of Global Agent is in centralized manner. Local Agents comprise of Load Agents

(LAGs) and Switch Agents (SAGs). A mathematical model based on graph theory is

developed for Fault detection in distribution network. Fault detection process is carried out

by the communication and coordination of LAGs. Each LAG communicates to its

neighboring agent and based on the local information of each LAG the fault location is

detected precisely. After the fault detection process LAGs report the fault location to GAG

in order to reconfigure and restore the network. GAG has its own database, where it

stores all scenarios and will recall these in future in order to speed up the reconfiguration

process.

The main objectives of the fault reconfiguration are, to supply the critical loads and to

maintain the good voltage profile across the network even in the event of faults. The fault

reconfiguration algorithm makes use of shunt compensation and priority based load

shedding in order to control the voltage. The fault reconfiguration algorithm proposed in

this thesis will choose the best possible switching configuration in regards to voltage

profile and real power loss in the system.

The MAS agent architecture is developed using JADE platform for fault detection and

reconfiguration. Global Agent will make use of MATLAB reconfiguration algorithm for

reconfiguration and restoration of power distribution system. All possible scenarios are

tested on the proto-type test system, Simplified CoF. The results obtained are very

promising and show superior ability of Multi-agent systems in the field of fault detection

and reconfiguration.

Fault detection and fault reconfiguration algorithms proposed in this thesis are very

efficient in terms of algorithm execution times. Execution times for both algorithms are

tested. It is observed that the average execution time taken by fault detection algorithm is

22.4 ms and fault reconfiguration algorithm is 1.8796 sec. When the same fault scenarios

ran from GAG’s database, a significant reduction in execution time is noted and the

average execution time recorded is 25 ms.


86

6.2 Future Work

MAS technology is a new area in the field of power engineering and has lot of potential

benefits when compared to traditional schemes. The technology is still in research arena,

and has greater number of possible extensions. Some of the future extensions to the work

cited in this thesis are listed below,

• The key feature of the Multi-agent system is its operation in decentralized manner.

In the proposed work, a centralized reconfiguration scheme is used by global

agent for circuit reconfiguration and restoration. Hence, a new decentralized agent

model for reconfiguration has to be developed.

• Interfacing agent platforms with the already existing power system software is a

main technical challenge, which exists even today. There should be a nice

interface, in order to apply the Multi-agent systems to the power system and to

harness its key benefits. In the proposed work, there is a need of interface

between JADE and Power World Simulator. This can be done by developing an

interface using C++ language. There is also a need of efficient interface between

JADE and MATLAB.

• The fault detection and fault reconfiguration algorithms are developed based on an

assumption of radial configuration of the system. In practical, most of the

distribution networks have mesh or loop configuration. Hence, a new set of

algorithms should be developed to accommodate all types of circuit topologies.

• The fault detection algorithm used in the proposed work makes use of Load agents

for fault detection. This algorithm has a disadvantage of not locating faulty links

instead it locates faulty sections. Hence, a fault detection algorithm based on node

agents should be developed.


87

• Finally, a complete MAS architecture should be developed which can cater all the

needs of power systems such as, fault detection, fault reconfiguration, voltage

control, protection co-ordination, incorporation of distributed generators, learning

and reactive power management.

88

REFERENCES

[1] http://www.eei.org/industry_issues/industry_overview_and_statistics/history

[2] Koushaly Nareshkumar, “Application of Multi-agents to Power Distribution

Systems” M. S. Thesis, West Virginia University, 2008

[3] P. Silpa, “Power Market Analysis Tool for Congestion Management”, M.S.Thesis,

West Virginia University, 2007

[4] http://en.wikipedia.org/wiki/2003_North_America_blackout

[5] Fabio Luigi Bellifemine, Giovanni Caire, Dominic Greenwood, “Developing Multi-

Agent Systems with JADE”, Wiley publisher, 2007

[6] G. Weiss, “Multi-agent Systems: A Modern Approach to Distributed Artificial

Intelligence,” The MIT Press, 2000

[7] M. Wooldridge, N. R. Jennings, “Intelligent agents: Theory and Practice,” The

Knowledge Engineering Review, Vol. 10, Issue 2, pp 115 – 152, 1995

[8] P. Tulpule, “Multi-agent Approach for Power System Reconfiguration,” M. S.

Thesis, West Virginia University, 2007

[9] S. Russell and P.Norvig. Artificial Intelligence: A modern approach. Printice-Hall,

1995

[10] F.Hayes-Roth, D.A. Waterman and D.B.Leat, editors. Building Expert Systems,

Addison-Wesley: Reading, M.A, 1983

References

89

[11] P.Jackson, Introduction to Expert Systems Addison-Wesley: Reading, M.A, 1986

[12] John R Searle. Speech Acts: An essay in the philosophy of Language. Cambridge

U. Press, 1970

[13] The Foundation for Physical Intelligent Agents, http://www.fipa.org

[14] FIPA 97 specifications, version 2.0, ACL, http://www.fipa.org

[15] D. A. Cartes and S. K. Srivastava, “Agent Applications and their future in Power

Industry,” IEEE Power Engineering Society General Meeting, pp 1 – 6, June 2007

[16] Z. Zhang, J. D. McCalley, V. Vishwanathan and V. Honavar, “Multi-agent System

Solutions for Distributed Computing, Communications and Data Integration Needs

in Power Industry,” IEEE Power Engineering Society General Meeting, Vol. 1, pp

45 – 49, June 2004

[17] L. Liu, K. P. Logan, D. A. Cartes and S. K. Srivastava, “Fault Detection,

Diagnostics and Prognostics: Software Agent Solutions,” IEEE Transactions on

Vehicular Technology, Vol. 56, No 4, pp 1613 – 1622, July 2007

[18] R. Isermann, “Process fault detection based on modeling and estimation methods

– A survey,” Automatica, Vol. 20, No. 4, pp 387 – 404, July 1984

[19] R. Isermann, “Model based fault detection and diagnosis – Status and

Applications,” Annual Rev. Control, Vol. 29, pp 71 – 85, 2005

[20] E. Chow and A. Willsky, “Analytical Redundancy and the design of robust failure

detection systems,” IEEE Transactions on Automatic Control, Vol. 29, No. 7, pp

603 – 614, July 1984

References

90

[21] R. J. Patton and J. Chen, “Advances in Fault Diagnosis using Analytical

Redundancy,” IEE Proceedings of Integrated Operations Management Control, pp

6/1 – 6/12, 1993

[22] K. Huang, S. K. Srivastava, D. A. Cartes and L. Liu, “Agent Solutions for Navy

Shipboard Power Systems,” IEEE International Conference on Systems

Engineering, pp 1 – 6, April 2007

[23] C. Angeli, “Online fault Detection Techniques for Technical Systems: A Survey,”

International Journal in Computer Science Applications, Vol. 1, No. 1, pp 12 – 30,

2004

[24] K. Huang, S. Srivastava and D. Cartes, “Decentralized Reconfiguration for Power

Systems using Multi Agent System,” 1st Annual IEEE Systems Conference, pp 1

– 6, April 2007

[25] J. M. Solanki, N. N. Schulz and W. Gao, “Reconfiguration for Restoration of Power

Systems using Multi-Agent System,” Proceedings of the 37th Annual North

American Power Symposium, pp 390 – 395, October 2005.

[26] J. M. Solanki, S. Khushalani and N. N. Schulz, “A Multi-Agent Solution to

Distribution Systems Restoration,” IEEE Transactions on Power Systems, Vol. 22,

No. 3, pp 1026 – 1034, August 2007

[27] T. Nagata, H. Sasaki and R. Yokoyama, “Power System Restoration by Joint

Usage of Expert System and Mathematical Programming Approach,” IEEE

Transactions on Power Systems, Vol. 10, No. 3, pp 1473 – 1479, August 1995

[28] T. Nagata and H. Sasaki, “A Multi-agent Approach to Power System Restoration,”

IEEE Transactions on Power System, Vol. 17, No. 2, pp 457 – 462, May 2002

[29] T. Nagata, N. Iba, Y. Tao and H. Sasaki, “A New Power System Restoration

Framework by means of Multi-agent Approach,” IEEE/PES Transmission and

Distribution Conference and Exhibition, Vol. 3, pp. 1695 – 1700, October 2002

References

91

[30] T. Nagata, Y. Tao and H. Fujita, “An Autonomous Agent for Power System

Restoration,” IEEE Power Engineering Society General Meeting, Vol. 1, pp 1069 –

1074, June 2004

[31] T. Nagata, Y. Tahara and H. Fujita, “An Agent Approach to Bulk Power System

Restoration,” IEEE Power Engineering Society General Meeting, Vol. 1, pp 599 –

604, June 2005

[32] T. Nagata, Y. Tao, Y. Tahara and T. Aoyama, “Development of Bulk Power

System Restoration Simulator by means of Multi-agent Approach,” The 47th IEEE

International Midwest Symposium on Circuits and Systems, Vol. 2, pp 337 – 340,

July 2004

[33] D. Liu, Y. Chen, G. Shen and Y. Fan, “Multi-agent based Approach for Modeling

and Simulation of Bulk Power System Restoration,” IEEE/PES Transmission and

Distribution Conference and Exhibition, pp 1- 6, 2005

[34] L. M. Tolbert, H. Qi and F. Z. Peng, “Scalable Multi-Agent System for Real-Time

Electric Power Management,” IEEE Power Engineering Society Summer Meeting,

Vol. 3, pp 1676 – 1679, July 2001

[35] T. Hiyama, D. Zuo and T. Funabashi, “Multi-agent based Control and Operation of

Distribution System with Dispersed Power Sources,” IEEE/PES Transmission and

Distribution Conference and Exhibition, Vol. 3, pp 2129 - 2133, 2002

[36] Y. Qudaih and T. Hiyama, “Reconfiguration of Power Distribution System using

Multi Agent and Hierarchical based Load following Operation with Energy

Capacitor System,” International Power Engineering Conference, pp. 223 – 227,

December 2007

[37] J. A. Hossack, J. Menal, S. D. J. McArthur and J. R. McDonald, “A Multi Agent

Architecture for Protection Engineering Diagnostic Assistance,” IEEE Transactions

on Power Systems, Vol. 18, No. 2, pp 639 - 647, May 2003

References

92

[38] J. Zhang and Q. Liu, “Application of a Multi-Agent Technology in Relaying

Protection,” International Conference on Power System Technology, pp 1 – 6,

October 2006

[39] D. V. Coury, J. S. Thorp, K. M. Hopkinson and K. P. Birman, “An Agent based

Current Differential Relay for use with a Utility Intranet,” IEEE Transactions on

Power Delivery, Vol. 17, No. 1, pp 47 – 53, January 2002

[40] S. J. Park and J. T. Lim, “Modeling and Control of Agent-based power protection

systems using supervisors,” IEE Proceedings on Control Theory and Applications,

Vol. 153, Issue 1, pp 92 – 98, January 2006

[41] Z. Xiangjun, K. K. Li, W. L. Chan and S. Sheng, “Multi-Agents based Protection for

Distributed Generation Systems,” Proceedings of the IEEE International

Conference on Electric Utility Deregulation, Restructuring and Power

Technologies, Vol. 1, pp 393 – 397, April 2004

[42] H. Wan, K. K. Li and K. P. Wong, “An Multi-Agent Approach to Protection Relay

Coordination with Distributed Generators in Industrial Power Distribution System,”

Conference Record of the Industry Applications 40th IAS Annual Meeting, Vol. 2,

pp 830 – 836, October 2005

[43] H. Wan, K. K. Li and K. P. Wong, “Multi Agent Application of Substation Protection

Coordination with Distributed Generators,” Proceedings of International

Conference on Future Power Systems, November 2005

[44] M. W. David and K. Sakurai, “Protecting the Electric Power Infrastructure with a

Mobile Agent Monitoring and Analysis System,” Proceedings of International

Conference on Information and Communication Technologies: From Theory to

Application, pp 497 – 498, April 2008

[45] W. Hu, L. Benedicenti and R. Paranjape, “Mobile Agent Network Management

System Performance Study in Frame Relay Network,” Proceedings of Canadian

Conference on Electrical and Computer Engineering, Vol. 3, pp 1499 – 1504, May

2002

References

93

[46] T. Shono, K. Sekiguchi, T. Tanaka and S. Katavarna, “A Remote Supervisory

System for a Power System Protection and Control Unit Applying Mobile Agent

Technology,” Proceedings of IEEE/PES Transmission and Distribution Conference

and Exhibition, Vol. 1, pp 148 – 153, October 2002

[47] H. F. Wang, “Multi-agent Co-ordination for the Secondary Voltage Control in

Power System Contingencies,” IEE Proceedings on Generation, Transmission and

Distribution, Vol. 148, Issue 1, pp 61 – 66, January 2001

[48] S. Gehao, J. Xiuceng and Z. Yi, “Optimal Coordination for Multi-agent based

Secondary Voltage Control In Power System,” IEEE/PES Transmission and

Distribution Conference and Exhibition, pp 1 – 6, 2005

[49] V. Krishna and V C Ramesh, “Intelligent Agents for Negotiations in Market Games,

Part 1: Model,” Proceedings of PICA97.

[50] V. Krishna and V C Ramesh, “Intelligent Agents for Negotiations in Market Games,

Part 2: Application,” Proceedings of PICA97.

[51] Hyungna Oh, Tim Mount, “On the First Spike in Summer”, Proceedings of the 37th

Hawaii International Conference on System Sciences – 2004

[52] Hyungna Oh, Tim Mount, “Testing the Effects of Holding Forward Contracts on the

Behavior of Suppliers in an Electricity Auction”, Proceedings of the 38th Hawaii

International Conference on System Sciences – 2005

[53] Junjie Sun, Leigh Testfatsion, “An Agent-Based Computational Laboratory for

Wholesale Power Market Design”, IEEE, 2007

[54] Yeung C S K, Poon A S Y, Wu F F, Game theoretical multi-agent modeling of

coalition formation for multilateral trades [C]. IEEE Transactions on Power

Systems, Volume 14, Issue 3, Aug 1999 929-934.

References

94

[55] Ping Wei, Yongle Yan, Yixin Ni, Yen J, Wu F F, Allocation of tieline costs in power

exchange scheduling using a multi-agent approach[C]. IEEE Power Engineering

Society Winter Meeting 2001.Volume 3, 2001 1220-1225.

[56] Reinprecht N B, Bones D, Data management for the electricity market[C]. IEEE

Computer applications in Power, Volume: 14, Issue: 1, Jan, 2001: 35-38.

[57] A. Aitouche and S. Hayat, “Multi-agent Model using Colored Petri Nets for the

Regulation Traffic of an Automated Highway,” IEEE Proceedings on Intelligent

Transportation Systems, Vol. 1, pp 37 – 42, 2003

[58] V. Gorodetsky, O. Karsaev, V. Kupin and V. Samoilov, “Agent based Air Traffic

Control in Airport Airspace,” IEEE/WIC/ACM International Conference on

Intelligent Agent Technology, pp 81 – 84, 2007

[59] K. Zafar, S. B. Qazi and A. R. Baig, “Mine Detection and Route Planning in Military

Warfare using Multi Agent System,” IEEE Proceedings of the 30th Annual

International Computer Software and Applications Conference, Vol. 2, pp 327 –

332, September 2006

[60] S. D. J. McArthur, E. M. Davidson, V. M. Catterson, A. L. Dimeas, N. D.

Hatziargyriou, F. Ponci and T. Funabashi, “Multi-Agent Systems for Power

Engineering Applications – Part I: Concepts, Approaches, and Technical

Challenges,” IEEE Transactions on Power Systems, Vol. 22, No. 4, pp 1743 –

1752, November 2007

[61] S. D. J. McArthur, E. M. Davidson, V. M. Catterson, A. L. Dimeas, N. D.

Hatziargyriou, F. Ponci and T. Funabashi, “Multi-Agent Systems for Power

Engineering Applications – Part II: Technologies, Standards, and Tools for Building

Multi-agent Systems,” IEEE Transactions on Power Systems, Vol. 22, No. 4, pp

1753 – 1759, November 2007

References

95

[62] G. Morejon, S. K. Srivastava and D. A. Cartes, “Integrating Virtual Test Bed and

JADE Agent Platform using CORBA for Reconfiguration of Shipboard Power

System,” IEEE Power Engineering Society General Meeting, June 2006

[63] The MathWorksTM , http://www.mathworks.com/products/matlab/description1.html

[64] MATPOWER home page, http://www.pserc.cornell.edu/matpower/

[65] PSERC PowerWeb project, http://www.pserc.cornell.edu/powerweb/

[66] Java Agent Development Framework, http://jade.tilab.com/

[67] F. Bellifemine, G. Caire and D. Greenwood, “Developing Multi-Agent Systems with

JADE,” John Wliey & Sons, Ltd., 2004

[68] H. Shaalan, J. Thompson, R. Broadwater, M. Ellis and H. Ng, “Distribution

Engineering Tool Features a Flexible Framework,” IEEE Transactions on

Computer Applications in Power, Vol. 8, Issue 3, pp 21 – 24, July 1995

[69] R. Broadwater, J. Thompson, M. Ellis, H. Ng, N. Singh and D. Loyd, “Application

Programmer Interface for the EPRI Distributed Engineering Workstation,”

Proceedings of IEEE Transactions on Power Systems, Vol. 10, No. 1, pp. 499 –

505, February 1995

[70] Power World Simulator, http://www.powerworld.com/Default.asp

[71] Wikipedia, http://en.wikipedia.org/wiki/Graph_theory

agent based modeling of power distribution systems

Documents